a business product, a project, rather.
We're gonna build a company in three days.
And so we're gonna start out with some intros,
a little bit of background, and I'll give you the run of show.
So with me, we're gonna do some intros here.
I have Lauren, hi, I'm Lauren, potato on Twitter.
Work on Graphbot at SpaceX AI.
Yeah, hey, I'm Roshan.
Work on Graphbot and a few other things at SpaceX AI.
Mostly focused on product and super excited.
Yeah, and I'm Matt Palmer,
working developer experience.
And the idea behind this live stream,
what we're gonna be doing for the next couple of days
is that we'll be working live from a studio in San Francisco,
right next to the Moscone Center at Dreamforce.
And we're gonna share every step of the process
in building a business.
We're gonna talk about how we would do this
using Graphbot, using the latest AI tools
from SpaceX AI.
And there are gonna be some mistakes.
It's not gonna be perfect.
We're gonna walk you through everything in the process.
We actually don't even know what idea we're gonna build yet.
We should probably figure that out.
We're still working on it.
We should figure it out.
But we're gonna get to it.
And through the three days,
we're gonna be mostly here in this room.
It might get a little messy, I don't know.
It's probably gonna get a little crazy.
But we're also gonna cut over to SpaceX team members
who are gonna share, number one, what Graphbot is,
how it works, and how you can use it for your job,
for your life, for everything really that you care about.
And you're gonna see a lot of that today
as we go about building our companies.
And that first session is gonna be Graphbot 101.
So if you know nothing about Graphbot,
we've worked with Roman and Rita.
They're excellent.
They'll walk you through everything from
how it got built to what it is today.
And I feel like Graphbot can be pretty intimidating
if you've never used it before.
It's a bit of a new concept.
Yeah, I think it'll be fun to see different folks
and how they use it in different ways.
I think we have a really cool lineup of people.
We got some engineers, we got some product people,
we got some sales people.
And so regardless of what you work on,
hopefully you get a feel for how Graphbot
could be part of your daily workflow.
But yeah, the 101 session I think is gonna be really cool.
It's awesome.
Yeah, I'm super curious to see everybody's workflow.
I mostly use Graphbot for personal automation
and some engineering work as well,
but never done it for GTM or market research,
or anything like that.
So I'd be curious to see.
Yeah, I feel like I always learn something too.
I mean, I don't know about you guys,
but every time I see someone else use it,
it's like, oh wait, I didn't know I could do that.
I didn't think of this particular thing.
Maybe we'll come up with some of those ourselves too.
That's why I hang around Lauren.
I just like talk about it.
Lauren, teach me how to use Graphbot.
And then I try and communicate that with you all.
So yeah, no, it'll be super exciting.
But before we dig in, I have my piece of paper here
with what I'm supposed to say.
There's a super important,
ultra important announcement I'm supposed to tell everybody.
You have a chance to win something.
So you wanna listen to this one.
So you're gonna have to watch along,
follow along in the live stream,
because our Graphbot Galaxy live stream challenge
is now live.
So we're gonna bring, I know this is exciting.
I hear woos from the studio.
So what we're doing is we're giving away
a trip to a starship launch.
So you're gonna be able to go to Starbase
in Austin, Texas.
And to your tour.
For yourself and to guests.
I wanna go.
Yeah, I can't think of anything
my girlfriend would be more excited than
for me to take her to a starship lodge.
So actually, and the runner's up.
So even if you don't win,
the runner's up get a tour of the SpaceX
rocket factory in Hawthorne.
In the Bay Area.
So you're gonna have to follow the atGrock accounts
and the atBot accounts.
And the entry will be by September 29th.
All the details of the contest are in there.
But what we want you to do is share
how you've integrated Grockbot into your work.
So you're gonna quote the official posts on X
with a description of your bot and a share template link.
All of those details are going to be
from the atGrock and the atBot accounts.
So you just go over there, check it out.
And again, you're gonna need a trip to go to Starbase.
We'll be pretty sick.
Really cool.
I would love to go, yeah.
We're like, we haven't even gone to Starbase.
I haven't gone yet.
Maybe next we can work on that.
Yeah, there's also some really cool footage
of a launch on the GrockBos that I see on X right now.
So yeah, check it out, x.com slash grock.
And you can see some more info there
on I guess how to enter and some details.
Awesome, awesome.
Okay, cool.
We got some time.
The main things we wanted to cover
were just kind of like the layout,
how this is all gonna work, the contest.
And now we can just start talking a little bit
about what we're planning to do.
And I also wanna give you guys a little bit
of more background on who we are,
like what we do day to day.
Because that can be ambiguous
and it's never really clearly laid out.
So our goal is to show you all basically
how AI can help you get more stuff done.
And more importantly, like how AI can free up your time
so that you're not bogged down
by all the menial tasks and the boring stuff
that we all have to do every day
that we've done for our entire lives, most of us, right?
Whether that's like digging through Excel sheets
or trying to figure out where that one message went
that you think somebody sent you last night
but you can't quite remember.
Or actually writing code and automating
some of the more manual parts of software.
But not just software, knowledge work,
making slides, making designs.
And the key message I think really
is that AI frees you up to do the stuff
you actually care about.
And so maybe we could just go around
to all tell you a little bit more about myself
while you do kind of a little intro and what I do.
And I would love for the rest of the team to share
because presumably if we're watching this,
you're gonna be spending some time with us
over the next seven to two hours.
Over the next seven to eight hours.
Trying to go from really nothing
to a fully functioning company.
And shortly we'll also lay out exactly
what we mean when we say company.
Because there's actually a lot of confusion on this online.
It's like, oh, they're building a business.
I saw some quotes, they're like, they're building a startup.
They're building a multi-billion dollar company.
We could maybe do all of that though, I don't know.
We could try, I mean, that's a try, yeah.
Someone told us we should build SpaceX.
So I was like, seven to two hours.
We already worked for SpaceX.
I feel like they got the rockets down.
Like maybe we could do something different.
But yeah, we'll get right into it.
And then again, my name is Matt Palmer.
I work in developer experience.
Some of my know developer relations.
I have some amazing teammates, Lee Rob, Eric, Milo,
just a great team.
And largely that looks like user education,
like teaching people how to use Rockback,
how to use these new technologies,
how to use these tools that really nobody's ever used
before, and some brand new things.
And so day to day, I'm making a lot of content.
But I'm also doing a little bit of engineering work.
Might be contributing code here or there,
updating our documentation, helping set up live streams.
Steven, maybe.
And so it's really a combination of content,
a combination of products, and talking to a lot of people.
Just being on Twitter and understanding
how the product works, how people are using the product,
all of these things.
And so I'm super excited to bring that into what
we're building today.
I'll probably be thinking high level about a lot of stuff.
We'll talk about individual roles as we move through this.
But yeah, that's a bit about me.
So maybe, Roshan, you can dig more into what you're
doing day to day.
I'm actually curious before we do that,
what's your favorite demo you've built recently?
Because you make some really awesome demos.
OK, cool.
Yeah.
So I shared this one yesterday, actually.
So it's on my profile.
Hopefully we have
our Twitter handles on the way.
Yeah, we should bring this in.
But Pat, Maddie P. That's all you got to know.
But I think this one was pretty cool.
I set up GrockBot basically to do the process of trying out
new technologies.
So it goes on my X bookmarks, finds a cool package
or a cool technology.
So I'll bookmark, oh, here's a cool MPM package
that can make maps or make a rotating globe.
And then it's like, OK, I have to find the time to sit down,
spin up that repo, actually have this thing running
to understand really what it is or read the documentation
or whatever.
And so this GrockBot looks on my X bookmarks
and finds things like that.
It will fire up a cursor cloud agent
to start building with that technology.
And then I configured the repo using GrockBot actually
to do preview deploys.
I think using Cloudflare or something.
And so every morning I just get a message from the GrockBot
and says, hey, here's this technology that was in your
bookmarks that you thought was cool.
And deployed version of it.
That's super cool.
Yeah, I have a walkthrough on my Twitter page
of how to do that.
So yeah, that was maybe my favorite.
I liked it a lot.
I'm glad I asked.
I feel like, again, this is like I always
learn something any time I talk to someone.
It's cooler than the stuff I do with GrockBot.
You've built GrockBot, or are you kidding me?
OK, well, I guess I'll go next.
I'm Roshan.
I work on product at SpaceX AI.
And I don't know, I think the product role has really
changed in the last few months and maybe year
plus where, I don't know, depending on which philosophy
you subscribe to, everyone is sort of a builder now.
But I don't know, I think a lot of what
I spend my time thinking about is how
to build the right things for customers
and how to kind of bring that customer feedback and what
people want out of their use of AI tools
kind of into the product.
And so in day to day, I'm thinking about things in GrockBot.
We've built a ton of really great connectors
into the GrockBot marketplace.
And so trying to get more and more of those connectors
in place.
And then also just thinking about how
we think about what to build next, like what
the next frontier ought to be.
And I'm actually kind of excited to get into some of this
today, because it's not often that you
get to start from scratch on a new business or a new entity.
Like oftentimes you're working on an existing platform
and it's like, what do we do next?
But I feel like what's going to be really cool here
is we're going to get to do a lot of the zero to one
stuff together.
So how do we go find data for what we should build?
Or what's the world saying about ideas?
And how do we go do some user research
and maybe go set up some interviews
and kind of run the whole product development process
and the whole product development playbook.
So yeah, I'm pretty excited for that.
I think it'll be pretty fun to get into the zero to one parts
of it.
And then of course, a lot of fun decisions
to make with the various, hopefully we'll make some bots.
But maybe we'll have a designer bot that
can help us design some cool stuff.
Or we'll have a marketing bot that
can help us market some cool stuff.
So yeah, excited to figure out how we coordinate all that.
I think it'll be a lot of fun.
Yeah.
Lauren, you want to go next?
Yes, I am Lauren.
I am Potato on Twitter.
And I use Grockfurt for a lot of really crazy things,
I think, really random things.
So a couple of weeks back, I tweeted out this kind of a joke
where I said, if you want to summon me on X,
just say Potato, Potato, Potato.
And then I will disappear and say something.
Of course, I have to automate that
because there's no way I could respond to so many mentions.
So I used Grockfurt for that.
Also do a lot of engineering work on Grockfurt.
So I think coming on the live stream
and talking to you all and working together,
it's going to be really fun.
Because usually I feel like I'm stuck in my cave,
my goblin cave at home, just coding.
I work on Grockfurt.
So I'm really obsessed about performance.
And so I'm just constantly just working on making the app faster
and faster and faster.
So I really love doing that.
We're all thankful for this.
Yeah.
And another thing I really like, or I spent a lot of time on,
is setting up the Grockfurt code base
so that anyone, whether you're a PM or a designer,
to contribute to the code and have it be good by default.
So I'm looking forward to setting something up
like that here too so that all three of us
can be building and iterating really quickly
together on what I don't even know what we're going to build.
But I think it'll be a lot of fun.
Yeah.
Oh, I'm excited for that.
Because I love the Grockfurt DevX.
Just coming in from the outside, it's
super easy to understand the way you laid things out.
We have channels where if there's an issue,
we just post what the issue is in some cloud agent somewhere.
And also usually you.
Like it's kind of shocking how fast you
appear when issues arise.
Starts working.
And I love that.
And so I mean today, when we're building,
we're going to use all the same tools you would
to start a company.
Like if you just sat down with three friends,
you'd probably like, I don't know,
set up somewhere to talk, set up somewhere to take notes,
and then start building using your favorite tools.
And so the cool thing about working at SpaceX is like, OK,
well, cool.
We have cursor.
We have Grockbot.
We're going to use cursor to write some code.
But then we're going to use Grockbot.
And we're going to figure out all the hard manual parts,
which is kind of like, OK, well, like, I don't know.
We probably want to sell something.
So we need to figure out what the idea is.
Well, yesterday we asked you all, and also SpaceX employees,
like, hey, what should we build?
Do we have any good ideas?
And so we have thousands of responses.
The first step is probably to ask Grockbot
to look at those responses, right?
It's to do market research.
So we're going to do some market research here
to kick it off.
And then we're probably just going to figure out
what we want to build, maybe announce that.
I guess probably this morning's good, right?
Yeah.
Well, we have three days.
It feels like we should have an idea in day one that
feels good to play at.
OK, cool.
Yeah.
So the goal is to have an idea in day one.
Maybe it's a good time to talk about exactly what we mean
when we say we're going to build a business.
Yeah.
So it's like, when you.
Yeah, what is a business?
I was going to ask you what a business.
I guess when you all think, OK, we're building a business.
Like, what does that mean to you?
Yeah, I don't know.
For me, it's like, I think you want to make something
useful.
I don't know.
I kind of go back to the build something people want
mentality, right?
I think we need to build something useful, something
that people want.
Maybe something we'd use ourselves.
You know, that's often a good way to, like,
I think we'll have a lot of empathy for what we're doing,
if it's something that we'd want to build and use ourselves
to.
What do you guys think?
I like the dog fooding part, because for me,
one of the things I really love about working on Grockba
is that I use it to work on itself.
So there's this really nice loop of being
able to dog food your own product and make it better.
You know, be the ultimate power user of your own thing.
So I feel like we definitely want to have something like that
for our business idea.
But yeah, when it comes to business,
I guess I don't know, I think it definitely
involves maybe making a little bit of money.
Oh, we should make money.
OK, I love doing it.
The team wants to make money.
It's not a real business until we've got to have it.
You know, see the dollar flowing into the bank account.
OK, so business dollars to bank accounts, building,
dog fooding, testing out the product,
building something people want.
Yeah, if I was going to add anything,
I would just say that I think it's easy to forget,
especially in like, we're all terminally online.
Like, if you're watching a stream, you're part of like X,
which is where people go.
They want to know stuff, right?
That's where X is where people go.
But that also means a lot of people are online a lot on X.
I am the biggest, I'm the most guilty of this.
I'm on Twitter, X way too much.
But what it means to build a business
is ultimately to serve humans, right?
And there are agents.
Well, serve humans are agents, potentially.
But the person asking the agent to do something
is human, right?
And so at the end of the day, it does kind of tie back
to like a human experience of like,
if I buy a t-shirt online and that gets delivered to me,
that might be a business idea, right?
But it's about the process of you
created this thing for a human being.
Or if you make Grockbot, we are
creating Grockbot for human beings
to interact with a computer, right?
And so it is very a very human process.
And it's something that at the end of the day is like,
how do we create something that you want as a viewer,
something that you find like magical,
or adds some sort of value to your life?
And so I think my favorite part is just like,
it really becomes about people and like,
how do you make a person happy?
Have you enough to like give you money?
It's like, can you do this again?
Or their agent.
Or their agent.
Yeah, I keep coming to this idea.
Like, in Grockbot, you can give your bot a credit card
now, like a link, a striped card, I guess, link card.
So I think it'll be fun to kind of play around
with some cool ideas.
Yeah.
What could we make that maybe a human could come in
and buy something, or maybe a bot could also
come in and buy something.
Exactly.
Maybe they're the same thing.
Maybe they're different.
Yeah.
It could be interesting.
Yeah.
That could be kind of fun too.
Maybe for all the people watching,
they'd want to like create a Grockbot account
if they don't already have one and like get their bots
to buy something.
That could be kind of interesting.
Yeah.
I mean, check it out.
We're not going to like, we're
just going to build the thing to this week
or these three days.
But you're more than welcome to follow along.
Yeah.
So I think the process of like human connection
is something we'll explore.
The process of like physicality?
Like, I don't know.
Do we build a digital product?
Do we build like a place where you can, I don't know,
go buy like coffee beans and have them delivered to you?
Do you like, or do we create something in person?
Like we're in San Francisco.
Like Dreamforce is going on right now.
There are like 50,000 people in the city here all.
It's mild.
Never seen so many people in that life.
50,000 additional people like in this block.
Like there's streets closed.
It was hard to get here.
There are a whole nine.
And so to me, it would be really cool
if we could like take something digital
and then use that to manifest something in the real world
that even you all can interact with.
And so like I think if I'm taking that away,
like that would 100% be my goal for this week.
But we don't know.
Like is that possible?
We'll be able to do it.
I'm just talking and like I feel like we're wasting time.
So yeah, I don't know.
I'm excited to see where it goes.
And like we're going to keep it super casual.
You're going to see us at our highs and at our lows.
It's going to be absolutely, hopefully,
entertaining, I don't know.
Yeah.
Yeah, I don't know.
I think that part of the fun is going to be,
we have three days, which is a lot of time,
but also not much time.
But I think we'll need to,
there'll be a bunch of stuff we have to set up.
Like I'm just starting to think ahead to,
we probably need a Slack channel for our team to chat in.
And maybe our bots to chat in too.
We need like a Notion space.
Yeah, we might Slack,
I mean what do we use that work?
Slack, awesome tool, Notion, awesome tool.
Maybe linear for tickets.
Yeah, we could check out a linear.
Yeah, or something like that.
Like do you use linear or how do you prioritize?
Like work?
I think we can try it,
but I'm also wondering if we can just,
because like if we have Slack,
like I could just like,
maybe as we use our own app and you report bugs,
I can just auto fix them.
So why open an issue when you can open a PR?
But I mean, yeah, maybe we'll get a backlog at some point
that we want to keep track of.
Okay, so maybe Slack is a control plan.
You know, like I wonder if we start to start with Slack first.
Yeah, Slack.
And then see what we need.
Slack, Slack, issues.
Issues, Slack, yeah.
Issues channel.
I can start wiring up my factory.
Okay, issue Slack channel.
We're gonna wire up the factory to the issues channel.
And then while we're doing that, like no matter what,
we're gonna start walking through the factory
because you're the infamous creator of P-Stack.
Right?
Famous.
Famous?
No, infamous, famous.
I don't know.
And for our viewers,
might not know what P-Stack is.
What is P-Stack?
So P-Stack is, it's a million dollar,
it's not, it's not, it's not, it costs nothing.
It's an open source plugin that I made
that takes all of the skills that I've built
for doing rigorous engineering work that I do all the time.
And I open sourced it so that anyone can reuse
the same skills and workflows that I have.
And I've been iterating on it for quite a while.
It took me quite a long time to get it to that point.
But last month I actually shipped
two and a half thousand PRs of P-Stack into production.
And I don't think I caused any bugs.
I hope I didn't.
So yeah, if you haven't tried out P-Stack yet,
you can go to the Grockbott Marketplace
and search for P-Stack and try it out
and let me know what you think.
I just did the math.
2,500 PRs in a month is 83 PRs a day.
And we have three days.
So, two times three is 250 PRs.
I think you can do more than that.
I can do more than that.
You should have a PR meter on the screen.
I have that track.
Actually, I'm not even going to open PRs.
I'm going to merge to main.
Oh yeah, there we go.
Yeah, yeah, yeah.
We'll let you know.
We'll call it a commit meter.
We'll have a commit meter somewhere on the screen.
We don't need pull requests.
Yeah, yeah.
Oh my gosh.
Yeah, I'm concerned.
The entire team is like, oh boy.
No.
And I feel like sometimes you say 2,000 PRs
and people are like, oh, that's just something
they say on stream.
But I physically have sat over your shoulder
and see you do some of this stuff.
And it's like, Grockbond, the app,
is built by people at SpaceX that are using tools like this
to rapidly iterate.
And so hopefully, we'll be able to show you that process
and we'll be able to show you that product as it goes
from quite literally nothing or ideating right now
to something that we're shipping to you all.
So yeah.
Yeah.
We were talking about tools before, right?
So it sounds like we're going to go Slack, Notion,
kind of stack.
We've got maybe linear.
If we decide we need that, we can just
start shipping directly to main.
I mean, we're going to be here talking to each other.
What else do you really need?
Anything else?
I mean, we have some database tech maybe.
Yeah, we need some like infra that we can set up.
We need the systems.
So we might need a database.
I know a database CEO lives around here.
They will have them on at some point.
Yeah, I can convince them.
Some of our friends to join.
Yeah.
We know we're big with the Stripe team, with Link,
with Payments.
So we need some payments in from.
Totally.
We'll probably start whiteboarding this at some point,
start building and showing people exactly where to do.
Oh, yeah.
Maybe we could actually do some actual whiteboard.
Yeah, we might get a chalkboard or whiteboard in here.
We've got the team.
That'd be super cool.
The team is listening to this now freaking out
because of our esoteric items that we're not
being building.
Please cut us a chalkboard.
On stream.
This is the great part.
It's like now that I'm on stream,
I can just ask for anything.
Presumably it'll disappear.
If I was like, there's a coffee shop down the road.
It's for the business.
I didn't get a chance to get a cappuccino this morning.
I'm wondering if in the next shot,
they'll just be like a cappuccino.
We can really push out far.
One thing I forgot that we're going to need a lot of tokens.
We're going to need a lot of tokens.
I think we have some.
I think we have a few.
I'm not talking about cappuccinos.
Yeah, we need tokens.
This is actually central to the idea.
Maybe you could tell us a little bit about what
the setup we're running is.
Yeah, so we're going to be using a combination of,
well, grokbot.
We're going to really try to drive everything with grokbot.
I think it's a plan here.
But we've set up ourselves on a grokbot team.
So we created a new team just for these three days,
where we're going to just use that.
We're going to each have our own teams of bots that we set up,
and hopefully be able to show you those bots as we go,
and be able to switch between them a little bit,
and let you interact with them a little bit.
And then, yeah, we really, I think we want to treat this
like a separate business and entity.
So I have a GitHub org that I made that's a completely
blank org.
Nothing in it, no repos.
So we can use that, if we like, as our starting point.
What's the name of the GitHub org?
It's called Ship by Thursday.
Ship by Thursday.
I love it.
Yeah.
Maybe we should open source some stuff too,
if people want to follow along.
It's just github.com slash shipbythursday.
It's completely empty right now.
So if you go there, you will see nothing.
But we'll need to put some stuff there.
Maybe we'll open source some Easter eggs.
I don't know, do you see?
No, I don't remember the org.
Actually, we literally haven't set any to stuff.
We didn't do any work beforehand, so none of this exists.
But we will start to make things available as possible.
And our goal is to make this as real as possible.
Yeah.
So we're starting with a completely blank slate, right?
At GitHub org, we have Grockbot accounts with zero bots in them
right now.
Brand new accounts.
Brand new accounts.
And so we'll walk you through from scratch,
how we'd install PSTAC to get set up for development.
Maybe in a few minutes, we can do some market research
like Matt was talking about.
I have some user research bots.
And we actually have a whole template of bots
in our marketplace too.
So we can find some useful ones there and pull them in.
I mean, we need to also hire a founding engineer,
probably.
A founding engineer and bot.
We just hire the bot.
We definitely need to hire a bot or two.
Two bots, intern bot?
I'm actually also wondering as part of market research,
do we want to have our bots actually kind of like the idea
you had where you have your Grockbot build something
for you every day?
Maybe we could have some quick throwaway prototypes
or something so we can visualize people's ideas,
something like that.
Oh, that's a really cool idea.
I feel like you have some bot factory stuff as well.
OK, the idea I'm talking about.
And this might be, if you're watching this,
you're like, Matt, I don't even know what Grockbot is.
Well, that's great because we've got a Grockbot 101 session.
We're coming up right for you in like five minutes.
But stick with me here.
You can create Grockbots.
Other Grockbots can create Grockbots.
It's wild.
Pretty crazy.
So what I'm thinking is, OK, at a meta point,
what do you do when you have a business?
You have onboarding.
So maybe we come up with a company doc
and that doc explains our process.
And we can also do that in Grockbot.
We could have a Grockbot that is a Grockbot factory
of some sort that creates our employees.
And so that onboarding bot might be like, hey, OK,
you want to read this doc, and that's who we are.
And that's what our principles are.
And that's what we're trying to do.
That little company handbook.
Like the state of the company.
And then we have to have a bot that updates the state.
Or all the bots update the state of the company.
Yeah, we could do something like that.
Share knowledge base.
Sure.
Right?
Most companies have some of that.
And so what I'm thinking is, as we add employees,
as we start to build out the team, OK, new, right,
we hire our intern bot, take a look at what we're doing in Notion.
And then we'll start prototyping.
But I like where your head's at.
I think when we resume back from the break,
we can start thinking about prototyping, market research.
But we need to start hiring bots.
So maybe in that 45 minutes, we'll
come up with a hiring plan.
Yeah, I like that.
We're going to need a bunch of bots, for sure.
And we're going to bring you all along for it.
We're not sharing our screens right now.
But we have everything hooked up, so you'll
be able to see exactly what we do.
You'll be able to see how we use the products in a clean demo
environment.
The one thing we did do before we got together
is we actually downloaded Grockwatt.
If you're starting a company, I guess that's the first thing.
It's the only thing you have to do, really.
And I'm excited.
So a little bit about the session that's coming up,
and some housekeeping stuff to recap.
If you're new to the stream, we're
going to build a company.
We've got 72 hours.
We haven't done anything.
I mean, we just spent the last 20 minutes yapping.
We shouldn't be waiting for too much.
We haven't done anything.
If you're watching, there is currently the Grockwatt Galaxy
livestream challenge going on.
If you share a template with us all and your template is
like the most.
It's like if a lot of people like your template, basically,
you have a chance to win a trip to Starbase.
Like watch a Starship launch, which is pretty sick.
Runners up get a trip to Hawthorne, where they
actually build some of this cool stuff, which is also
great.
And yeah, when we think about business,
we think about something that people want that creates value
for people.
That's where we're going to try to build.
With a human touch.
With a human touch, most importantly.
That's something we're going to try to build this week.
We're going to bring in different guests.
We're going to bring on guests of companies that we
like, guests of people on the internet that we like.
We're going to bring in SpaceX employees that are
really great.
They're all really great.
I don't pick favorites.
But we're going to bring in some especially great
employees that are going to show us their workflows.
We're just going to be hanging out.
They're going to help us.
They're going to power up.
We're going to ask for advice.
We're going to seek the knowledge of the crowd.
No man is an island.
Like, no entrepreneur ever did anything by themselves.
They had people to help them out.
So we're going to call on our friends to help us out
and give us advice.
But it's going to be us with the laptops here.
And there might be a few moments where we're just
furiously typing inside.
It's on stream too.
I feel like we're going to need some lockdown.
There will be some.
Hey, look, if you've watched any of the great streamers,
shout out.
Prime big friends of ours.
They've helped us out with some of this stuff.
It's going to be like that.
We're just going to be hanging out.
This might get a little bit messy.
The production crew is going to yell at me.
Make me clean the table up or something.
But we're going to have fun along the way.
Anything else you guys want to add for the viewers?
No.
It's going to be a lot of fun.
So I'm looking forward to building this with you all.
Awesome.
And I'm like, yeah, definitely keep shouting at us on X
if you're wanting to send us feedback.
And there's an active chat happening in the stream too.
So please drop stuff in chat.
We're going to get a setup so we
can see some of those comments rolling in and interact
with some of those suggestions in real time too.
So it'll be really helpful to us as we're steering.
And we really want to get folks feedback
and make sure this is useful to you.
So we just definitely encourage you to keep the feedback coming.
Three of us are all on X. I think our handles
are all over the place, hopefully by now.
So please post at us or reply to us.
And we'd be happy to bring some of that in as well.
But yeah, I see we have about 80,000-dish views
right now.
And so yeah, a lot of people out there who are really
excited, and we're really excited to hear from you
and ask for your help in crafting what we do here.
Yeah.
And one thing, I definitely recommend installing Grockbot
if you haven't already, signing up.
We have a free trial.
Free trial.
So you can try it out and maybe follow along with us
as we build and show you some of our pro tips.
We're going to tell you exactly how we do it.
So if you just download Grockbot,
we're going to give you the plugins, the templates,
all that stuff.
And most importantly, stream.
We're going continuously live from 8.30 to 6.30.
You'll have to check the schedule.
Tune in here anytime.
We're going to cut to workshops.
So if you download Grockbot, if you install the product,
you're going to be able to follow along with exactly how
it works.
If you were like, oh man, I want to learn Grockbot,
but I have to pay for this course.
No courses, no payment, all free.
Just watch this stream.
This is the course.
Someone told us, I got 25 seconds before we go live,
but someone told us their boss said,
you have to watch this live stream all 30 hours.
I'm sorry.
You're probably watching right now.
I'm sorry about that.
But you're going to learn a lot.
We're going to go to Grockbot 101 with Roman and Amrita,
two amazing, intelligent, beautiful, smart, high-achieving
functioning people that actually built Grockbot.
So we're going to kick it over there.
And they're going to teach you a little bit about Grockbot.
Are we good?
Whoo.
Well, you know, that's a OK.
We got the nurse.
Shake it out.
I'm going to teach you.

So first I wanted to talk a little bit about how Grockbot came to be.
And I think we're in a pretty interesting moment right now where when you think about
the way that AI has changed and really transformed the way that people and companies do work,
I think we've been through a few different phases.
And I think it started with really chat where you ask a question to AI, AI responds,
and that was really useful.
I think that helps with a bunch of research workflows, for example.
I think the next step after that was you delegate a task to AI.
And there's a co-pilot that kind of sits next to you while you do your work.
You iterate on a slide deck together, those types of workflows.
But I think we're really on the cusp of this new moment in AI where AI
starts to act more like a teammate or more like a colleague inside of a company.
And that's really different, I think, from what's come before.
And I think it's going to change the types of products that really meet the moment.
I think the ways that we need to change our ways and how we do work day to day
to really maximally get the most out of what AI can do inside of a company.
And then I think what will come after that is kind of this weird world of AI
teammates and human teammates that collaborate together inside of companies.
And where you're not just delegating a discrete task and saying,
hey, go from point A to point B, you're actually doing a lot more
kind of what you would do with an actual human colleague of they're not a collection
of tasks that you kind of outsource, but they own outcomes and they have responsibilities.
And you trust them to kind of own a discrete part of a company and kind
of do that work for you.
And I think this is really where Grockba originated from.
And I want to talk a little bit about some of the product decisions that we
made to get there.
So if you haven't used Grockba, this is what it looks like.
And I think there were a few things that we chose to do with Grockba that are a
bit different from other AI tools on the market that you maybe have used
before or familiar with.
I think the first one is we really wanted Grockba to feel a lot like
you're working with teammates and colleagues.
And so as a result, it feels kind of simple and chat-like and
almost like you're messaging your coworkers or teams of colleagues.
So you have on the left all of these bots, in this case, sales outbound and
an inbox manager that each have specific roles and different kind of lanes of work.
And the idea is not that you create a chat for each task that you want to take on.
You actually create a bot for a unit of work and a kind of job responsibility.
And you come back to that bot over and over and over again and it learns from you.
And I think that's one of the most magical moments of working with Grockba is
you say something a few weeks ago about, I really like when my slides are
formatted in this particular way.
And then Grockba remembers.
It remembers weeks later when you're making your next slide deck.
And that's what you would expect from a colleague and
I think that's what you should expect from AI as well.
So there's this teammate paradigm instead of a task paradigm.
Team mates really sit at the center of Grockba and the way to work with AI.
The second thing that's a little bit different is Grockba,
similar to what you would expect from a colleague,
doesn't just have access to direct integrations and MCPs and APIs.
Those are very useful.
They oftentimes don't get you 100% of the way to completing the work that you want.
They have access to their own computer and
they can use that computer exactly how you would.
They can watch a video, they can listen to a podcast, they can click in some
antiquated government software from the 90s.
We can do everything that a human can do with a computer, but it's yours.
And in addition to all of the integrations and MCPs and
things that you come to know and love with AI, I think it really opens up
a new set of tasks and opportunities for what AI can take off of your plate.
And I think it feels really different when you work with AI that can do those
things versus can't.
And the difference between getting kind of 90% of the way there of
almost having the job be done versus the thing just being done and to end kind of
hands off, you trust that your bot can do it, feels very different,
feels really magical.
And then the last thing is everything runs in the cloud.
It doesn't run on your computer.
And I think there's kind of in this weird moment in AI right now where a lot
of the workflows that people rely so much on AI for,
it runs on their local laptop, which is kind of strange.
And so if you're like asleep and your computer's closed,
the automation that you set up to run it to AM doesn't fire and you wake up in
the morning and you're confused and you want to kick something off from your
phone but it's tethered to your laptop back in your office.
It's just all this weirdness.
And we wanted to just jump right past that to a future where you have these
very smart colleagues, these teammates, these spots that live in the cloud and
you can interact with them anywhere you work.
It might be on your phone, it might be on your laptop.
In the future, it might be something entirely different.
And that really informed all of the decisions we made for the product was
those three things.
Feels like interacting with teammates, everything runs in the cloud.
And your bots have access to a computer that they can drive and
manipulate just like you can.
And I think part of the reason we really felt the desire and
the excitement to build something like RockBot is from what we've seen in
coding.
And I think from the past two years,
we've seen the job of a developer completely transform.
And I think a very stark, hard to overestimate how big the change has
been from an era of you're writing individual lines of code by hand to,
I think, what a lot of development looks like now, certainly inside of SpaceX AI,
where you have developers that own outcomes, they own direction.
And they steer their team of agents to complete these discrete tasks and
do kind of a lot of the mundane work for them.
And we haven't seen that for knowledge work at all.
I think when you look at the way that knowledge work is done today versus
the way it's done in 2024 or 2025, it's not that different.
And I think our hope and goal for
RockBot is to really be the most useful way to harness these models inside of
a company to do real work.
And so I touched on some of these points, but Amrita will talk a little bit
more about specific ways that we're seeing RockBot be useful for us
internally, but also inside of real companies in the wild.
But we wanted it to be delightfully simple.
And we think the future of AI products really should be very simple to use,
very simple to get started on, and then gradually pull you into
the more complex capabilities.
And I think a lot of people's experience using RockBot is you meet it on
its face, you kind of give it an initial task that is a little straightforward and
simple.
RockBot does it, you give it something more ambitious.
RockBot does it, you give it something even more ambitious.
And you just kind of keep going through that process until you feel the limits of
maybe where current AI capabilities are.
And our bet is these models are gonna keep getting smarter.
We want to give them a safe environment to run in to really maximally
be useful to employees of companies.
So it's very simple, it's always on.
It's this teammate in the cloud that you can always depend on and
rely on and never goes down even when you're asleep or
when your computer is closed.
And it really finishes the work.
You give it outcomes, it delivers.
And I think the next step of this will be how this works inside of teams.
I think right now AI is largely a single player game.
You work with your personal AI assistant on various things.
I think we're really excited about what shared AI colleagues could look like and
what the future of internal tooling even becomes.
So with that, very excited to pass it off to Amrita to talk a bit more
about concrete ways that people are using GrockBot and ways that we're using GrockBot.
Amazing, thank you, Roman.
So you'll see on the agenda and also for the rest of the week we have a lot of
sessions about how to use GrockBot at the enterprise level.
So you'll see GrockBot for GTM, for engineering, for marketing, and
then for admins.
Today's session is gonna take one of these actually and
then do a live demo with you all so you can see exactly what Roman mentioned.
Which is agents acting as colleagues, agents using their own computer.
And then also having memory and
an ability to learn from what you've taught them.
One thing I do wanna highlight is if you haven't registered for
any of the sessions already, please go on to I think x.ai slash galaxy.
And check out what's available this week.
We have a lot of really great sessions.
Basically about how to use GrockBot for
all of these different use cases within an enterprise.
Okay.
So now we're gonna go into the live demo session of today.
I do wanna ask, who in the room has used GrockBot?
Okay, looks like a good amount.
So maybe some of you will be seeing it for the first time.
But these are the three bots that we're gonna work with today.
We have Data Dan, Slide Sonya, and Email Ethan.
And they're gonna work together on a project that will actually require
you guys to participate.
Don't worry, no one's gonna have to speak or anything.
But what it's gonna do basically is collect user data.
Use that data to create slides and charts, and
then actually email that data to a stakeholder and external stakeholder.
And so what you'll see actually is how all of these agents will work together.
And to Roman's point, use tools that might not have native MCPs or APIs.
But make it very easy for you to kind of outsource this work so
that you can focus on more pressing things at hand.
All right, so I'm gonna minimize this and
we're gonna open up GrockBot.
And you'll notice that it looks very similar to maybe iMessage or
a tool that you're familiar with.
I have two bots that are already ready on the side, Email Ethan and
Slide Sonya.
And I'll talk a little bit about them later.
But what I wanna do with you all today is I wanna create a bot from scratch.
So you all can kind of see that workflow.
And this is the bot we're gonna create is Data Dan.
So I'm gonna call Data Dan.
I'm just gonna create him there.
And when you create a bot,
actually what GrockBot will do is it will give some idea of maybe what you wanna
use the bot for just based on the name.
So if you create a bot called Workflow Builder,
it'll automatically kind of have an idea of what it's meant to do.
You'll notice here that it gives some options of what do you mainly want me for,
which is great.
But I'm gonna start by actually asking it to create a Google Form for me.
And the reason I choose Google Form is, first of all,
I don't know if anyone likes making Google Forms.
I certainly don't.
But it's the most common way to get a lot of data from your employees,
from your customers, from if you're doing user research.
You can do this similarly with Qualtrics or
any other kind of form based software.
You can ask GrockBot to basically create it for you.
So what we're gonna do is I'm just gonna say, and
I'm gonna use voice mode here actually, so we can kind of see what that looks
like.
Create a Google Form for me that asks two questions.
The first is how many cups of coffee do you drink a day?
And the second is what are your favorite coffee shops in San Francisco?
And for the coffee shop options, maybe put ritual,
fills, sight glass, and blue bottle.
So it's gonna transcribe that.
What's really cool about voice mode, and
this is actually deployed internally right now,
is we actually have a back and forth voice mode where Grock voice can
actually talk back to you.
That should be out hopefully sometime within the next week or so.
But this is kind of a good way to see that speech to text in motion here.
So you'll say, you'll see that it kind of took everything I said.
I'm actually gonna move to the side here so you all can see.
And then immediately when you send a request like this,
GrockBot is going to respond and immediately start working.
And when it works, for example, in something like a Google Form,
what it will do is it will actually start using its own computer.
So at the top right is where you see access to GrockBot's computer here at
the top.
I can actually click into it and see what it's doing.
Give it a little moment to connect.
But once it's connected, you'll see that it has access to its own Linux VM machine.
So I'll be able to log in to Google Slides, Google Forms, Google Docs,
whatever you need.
And so what you can see actually, if I minimize this,
is it's gonna build both of those questions for me, and
I can monitor how it's building those in those forms.
So I mentioned Google Forms as a way to kind of interact with tools that
don't necessarily have MCPs or APIs.
But what I also wanna show is the built-in MCPs that we have already.
So you can see here that I have a bunch of MCPs or
plugins that I can connect to my GrockBot.
And depending on your task at hand, you can always manage these here.
And so this is kind of a good way to see what existing APIs and
MCPs are there already, or you can ask GrockBot to set it up for you.
Okay, so it looks like the bot wants to run a task for me.
This is a good example of how bots ask for approval.
I'm just gonna allow this for now, but
I'll talk a little bit about bot approvals later.
And how you can actually set particular rules saying, hey,
I only want a bot to send an email after it asks me for permission.
Or I only want the bot to send a Slack message to an external customer
after I approve it.
And this kind of makes it a lot easier for enterprises to control what the bot
can or cannot do in a big setting like this.
So this is gonna work on the Google Form.
While it's working, let me show you the other two agents that we'll be working with
today.
So email Ethan and Slide Sonia.
You'll notice that Slide Sonia already has the slides that I'm gonna be editing
up on her computer, similar to what Data Dan will be actually.
And what's nice about Slide Sonia is I'm able to basically edit the slides
myself or I can actually teach her how to do a particular task.
So you'll see here at the top, teach a task.
I can actually show Slide Sonia, for example.
If I wanna create a slide that takes all the data from the Google Form and
adds an animation to it or makes it a particular color,
I can actually show her how to do that.
So if I hit Teach a Task here, I can actually say, let's say I wanna just
teach her how to do an animation.
I can hit this, I can go to Insert and I can add animation here.
And notice that this is me controlling the bot's computer.
So this is not the bot doing this on its own,
this is me kind of controlling it.
And I can say, okay, this is how you would do a kind of,
let's do a fly in animation right there.
And let's just play it to see how it works and make sure it works okay.
Give it a second.
Cool, looks like it's monitoring the time that the animation takes.
Amazing.
So now that it's set up here and I knows exactly how to do it,
I can stop the task.
And immediately now that I've stopped the task and
Slide Sonia knows exactly what to do with an animation.
It's able to turn that recording and that animation into a skill.
And so now anytime I wanna say, hey, I want you to add that animation that I showed
you to a particular text or to a particular image on a slide,
it'll be able to do that from scratch.
Now go ahead and look up my data Dan and see how they're doing with
the Google form.
You'll notice that what's nice about, I'm gonna go ahead and
allow this again.
What's nice about having these agents is that I can always go back and
forth between them and also see how they operate within each other.
So as you can see, Slide Sonia is learning from the skill I just taught her.
Data Dan is working on the Google form.
I can actually see him creating the form live for me.
And then what's nice about this is once the form is created,
I'm actually gonna ask Data Dan to turn it into a QR code and
have you all scan it and fill out those two questions.
So on average, how many coffee cups do you drink a day?
And your favorite coffee shop in San Francisco?
And then we'll take all of that data, give it to Sonia, and
then have email Ethan kind of consolidate all that analysis and
send it out to let's see who the top coffee shop owner is.
And this is kind of a good way to see how you can take a very simple task
that usually takes a lot of time.
For example, an event planner at a company who needs to handle dietary
restrictions for a 10,000 person on site.
Or a user researcher that wants to collect data from all of their different
customers who are using Grockbot or any other tool.
How you can basically do all of these tools within Grockbot and
then have it kind of orchestrate that work for me.
I'm gonna go ahead and sit this one more time.
While this is working, let me actually show you the approval situation,
since we are seeing it here.
You'll notice that in the settings, when you actually scroll down here,
you will see a particular set of rules that you couldn't allow
Grockbot to basically do with or without your permission.
So what I'll actually show it is I'll say something like, hey,
I don't want you to reply emails for me unless you explicitly ask for
my permission first.
Or similarly, I want you to be able to create slides for me
automatically without asking for my permission.
And so that makes it a lot easier for
you to kind of granularly control what Grockbot can or cannot do.
Especially when it's doing long running tasks that you want to be
a little bit aware of or kind of cognizant of what it's doing.
So I'll give this a little more time while it's working.
One thing that I do want to say about Grockbot and how basically it's able to
run is that the computers that it's using on are all isolated.
Meaning that one Grockbot cannot access another Grockbot's computer.
So the fact that Grockbot is working on this Google form here is not
gonna impact the fact that Slide Sonya is working on the slides over here.
And that makes it easy actually for
one Grockbot to talk to another and work on two different things at the same time.
So a way that I use this is I often edit the same set of slides that my Grockbot
is editing.
And that makes it really easy cuz I can have Slide Sonya edit slide one or
two of mine of my slides and I can edit the 10th or 11th slide.
And we're not actually conflicting or making changes.
We're both kind of collaborating on the same set of tasks.
And that's kind of at the end of the day what Grockbot is.
It's a colleague that you can collaborate on.
What's nice about Slide Sonya is in her description here,
you can see that I've given her very explicit instructions on how to edit
an existing slide deck.
I can tell her that you're supposed to be using a particular font with
particular colors.
I actually always ask her to take a screenshot of the slide that she's
created once she's done and send it back to me.
And this kind of makes it a lot easier for
me to go back and forth with Slide Sonya because she already has this
predefined set of notions that she needs to get done.
And so I like doing this or I like giving my bot very explicit instructions
because that makes it more consistent.
And then if you ever want to add skills or routines,
you can always add that to your bot as well.
So while Data Dana is working, I can add a routine to Sonya that says,
every morning at 9 AM, I want you to summarize for
me any changes that were made to the slide deck.
And let me know who made those changes.
So this is a slide deck that I've shared with a ton of my colleagues and I want
to know if someone took out a slide, added a new part to a different slide.
I can ask Sonya to actually keep all of that context and then set up a routine
for me that will send that information to me every day at 9 AM.
And you can set up routines across multiple bots.
So you can have a routine that says, hey, Slide Sonya, email Ethan,
can you guys work together and actually email a summary of all the changes
that were made in this slide deck to my boss once a week.
And so he is aware of what's going on there.
And that makes it, again, a lot more collaborative and
a lot easier to work across these multiple agents.
Cool.
I'm gonna open up Data Dana again and see how the form is going.
Looks like it's making some changes.
It's asking for a lot of approvals today, which is funny.
I think it's because maybe it knows I'm on camera.
But what it's gonna do is once it's done kind of making this form and
designing what we should put inside of the questions, what it's gonna do is it's
actually going to take this form and turn it into a QR code that I mentioned you
all will scan.
And then what's nice about that is you no longer have to send out links for
forms.
I actually always just use QR codes that are made by Grockbot for
any form of external facing links or URLs that I have to show.
And then it will take basically all of this information and
put it inside of a Google sheet that again it can access on its own computer.
So in fact what I can do is while it's working on this,
I can go ahead and just actually look and
see at the other forms or sheets that I have made inside of Grockbot itself.
And so this is again me using Grockbot's computer to get access to
the forms or the sheets or responses that it's made.
And what's nice about this is sometimes if I wanna take control over
what it's doing or if I feel like it's taking a long time on a particular thing,
I can go ahead and just do it on my own.
And then again, teach it how to do it in case it takes a long time or
in case it's not aware of exactly what it should be doing.
So what it's gonna do is it's gonna go ahead and
make these changes a little bit more.
And then while it's working,
I'm actually gonna talk about email Ethan and
how we actually use email very, very usefully at Grockbot or
at cursor but with Grockbot's integrations with Google Drive and
Google Docs.
So email Ethan's job is basically to take all the data that slide Sonya and
Data Dan took and basically consolidate that and
send it out to external stakeholders.
So what I like to do actually is I tell email Ethan to look at real
existing emails that I've written before and consolidate that information into
a particular skill or type of tone that it can use for further emails.
So for example, I could say something like draft an email to,
I'll say Jason DeMore, someone in my context,
about the data that I've gathered on coffee habits.
And since it's still working on that data,
it's not gonna get anything just yet but
I'll show you all what drafting an email actually looks like in Grockbot.
What it will do is it'll actually be able to pull in data from either my own
personal source or from sheets that I have.
And then set up a little modal inside of Grockbot that actually asks
the user, hey, can I get your data from, or hey, can I send user?
As I'm, so this is actually where I would say something like when I want
Grockbot to reply to emails for me, I'm actually going to ask first and
add that as a rule.
And then now what it's gonna do is before it sends any email,
it's always gonna ask me first.
And that makes it so that I don't have to always monitor or
worry about what it's gonna send to any external phasing customers.
Cool, okay, so one thing that I wanna show is the agents talking to each other.
And so what it'll do actually is I'll tell email Ethan while it's working,
and this is also an example of bot steering, of how it can talk to
Slide Sonia and get information from what she knows and
context from what she knows.
So there's many ways you can do this.
You can always ask your bot in its description to delegate to other agents.
So for example, I think some people I've talked to have a chief of staff agent.
In your description, you can say something like,
like I always want you to delegate to your team.
And that way the chief of staff actually doesn't do anything on its own.
It delegates to its subordinate agents that have the right context.
But what I'll do here instead is I'll say something like you can see here at
Message Data Dan, completely on its own.
I can say something like can you message Slide Guru or Slide Sonia?
And ask her for any details about the GROC Bot Galaxy 101 slide deck that she made.
I want to send an email to my coworker about what we showed today.
So you'll notice that it was smart enough to know to Message Data Dan for
the coffee related information.
But now I'm explicitly tagging or calling another agent that it can work with
in order to kind of get the right context and information at the right time.
And that's what kind of makes these agents so powerful is they're experts and
they're personas in their own field.
And so you'll see in kind of the GROC Bot for engineering and
GROC Bot for sales demos is those agents are long lived personas.
They have memory.
And so because of that, they're not just scope to one task or one project.
They can be used for multiple tasks and multiple projects.
I can have email, Ethan work for the next year if I wanted.
And now he will have context about every email that I've sent.
And he'll be able to learn from all of that.
Similarly, Slide Sonia, she doesn't know just how to make this one type of slide.
I can teach her how to work in a different branding or a different theme.
Her job is just to create slides generally.
And so I can always change the scope and the persona of these agents in order
to make them as useful to me as possible.
The way I like to think about GROC Bot generally is you're basically creating
your own company.
You're creating your own team of agents and
you have the choice and the decision of how you wanna orchestrate that.
If you want your agents to be divided up by project, you can do that.
If you want them to be divided up by what they're expert in, you can do that as well.
And so it's kind of up to you and how you wanna orchestrate them and
have them work with each other on particular tasks.
So as you can see, I'll show you both how email Ethan messaged Dan and
how he also messaged Slide Sonia.
So you'll see here, these are the messages between GROC Bots.
You'll notice that email Ethan will always say,
Amrita asked me to do this thing.
Data Dan will say, I don't have this right yet.
I'm still working on the spreadsheet.
And email Ethan will say, okay, I'll use my existing coffee cup sample
data set that I have, amazing.
And then if we go back actually to, if I wanna go back to email Ethan here.
Go back to what he messaged Slide Sonia, you'll see kind of same thing.
Amrita needs details on this.
Please come back with a concise briefing deck about this and
give me any more information that I can use going forward.
And then you'll see that Slide Sonia has that same message from email Ethan and
is able to kind of respond back and see what's going on here.
So if I click on her again, you'll see that, okay, he asked for
this one-on-one briefing for a coworker.
I'm gonna scan the full deck and then pass Ethan an outline when it's ready.
So this is agent to agent protocol, agent to agent communication,
which makes it really easy for these agents to work together.
But what I also like about it is that I no longer have to monitor
what's going on.
I can kind of be more of a manager and overseer and
make sure that everything is going well.
But I don't have to get into the nitty gritty details of what these agents are
doing in order to kind of get the job done.
As Roman mentioned, the goal of Grockbot is to have finished work,
whether that's slides being made, emails being drafted, forms being created.
But at the end of the day, I am just making sure that the orchestration
is working rather than kind of going in and seeing exactly what's being done
here.
Okay.
All right.
So I'll say this coworker's for Jason, that's fine.
And then what I'll do is I'll check on my form one more time and
see how it's doing here.
As you can see, it's still messaging with email Ethan, but
now it's worked on creating this How Many Cups of Coffee, Do You Drink a Day,
and What Are Your Favorite Cops in San Francisco.
So it's working on kind of adding those options here.
The speed of these kind of computers really depends a lot of times on,
again, the model and how we're working on computer use as a company and as org.
You'll notice that computer use is going to become a very, very hot topic for
Grockbot, for Cursor and SpaceX generally.
And you'll see computer use continue to be improved over the next couple weeks.
You'll see actually what I really like about these machines is because they're
Linux VMs, they're very lightweight and they're very easy to spin up.
And so what I actually do sometimes is if I'm on an airplane and
I have bad Wi-Fi, I actually outsource all my agent work,
my computer use to my Grockbot because the Wi-Fi on its own Linux VMs is really,
really great, it's very fast.
And so this is kind of how I do any outsourcing of work, again,
for things that don't have MCPs, so Google Forms, Qualtrics.
And then once it starts adding options like it's doing here,
it makes it a lot easier for me to kind of create these really long running
tasks without me having to manually go and do this.
I can again focus my work on something else.
Okay, all right.
So as you can see, the bot always asks for
permission on connecting to outside services.
So as you can see here, it says the agent is trying to reply to an email.
And this is blocked by my own personal auto review rule.
So it's respecting the rule that I put in already, which says that,
hey, I don't want you to reply or I don't want to reply to emails for me.
So I'm going to go ahead and just say allow once so that it's able to kind of send
that content over.
And then what it will do is it'll actually show me a draft of that email.
What's nice about these agents is let's say I want to do this maybe twice.

three times, four times, you can always duplicate an agent and have maybe like five or six email
agents and sometimes I'll do this if I'm emailing different stakeholders.
So I'll have one email agent that's designed for emailing, you know, execs or CTOs or
CFOs.
I'll have another one that's designed for emailing just general users of GROCBOT,
you know, or, you know, regular folks who are using it every day to get feedback.
I'll have another email agent that's meant for, you know, messaging my team at work.
And so they all are tailored a little bit differently to be able to understand how
different, you know, stakeholders will respond to a particular type of email.
And so this is really great if you ever want to duplicate or try multiple
different strategies in an agent.
And then one thing I also love to show is sharing.
And so what you can do with a bot is you can always share it as a template
and then give it to other people in your team.
And this is what really kind of uplifts the multiplayer aspect of GROCBOT,
is you can have one agent that's shared across your whole team as a template
and then people can customize it for their own personal use case.
You might have seen that earlier actually when I hit Marketplace.
You can see here that these are featured bots, these are actually just public.
So Lauren is from Cursor and this is her bot that she's published.
But Clairvaux, Lenny and Eric is also from Cursor.
These are all bots that they've published that I can eventually just use
and incorporate into my own GROCBOT ecosystem.
And then you'll notice that these are, you know, from my team itself.
So if I ever want to publish my own GROCBOT to my team, to my company,
I can do that as well.
And it makes it really easy as I mentioned to have bots that are similar
so that there's consistency.
For example, if you are on a design team at work and you have a brand design bot,
you want to make sure that the brand design bot is following the same colors,
fonts, typography as everyone else's.
And so you would publish one brand design bot
that everyone at your company can then use.
And so publishing bots like this makes it really easy to have consistency.
But then it's also really easy to kind of, you know, modify them,
change them, and then people will get the most updated version of your bot.
So if you haven't shared a bot yet, please do.
It's really, really easy to do.
You just hit this again, as I mentioned, share as template at the top.
And then anyone will be able to kind of see your bot and use it for all its glory.
Okay, so here's an example of how an email draft looks inside of GROCBOT.
You'll notice that it has this really nice modal here,
which basically allows you, again, to edit things, change things,
do whatever you want.
I can even change, you know, who my recipient is,
if it's not the right person that I wanted to send it to.
I can also delete it or ask it to rephrase it
even if I don't like the writing.
And so this is kind of a good way of doing any form of email drafting within GROCBOT
so that you don't necessarily have to comb through all your drafts inside of your emails.
You can kind of do all of that within GROCBOT itself.
And so we do a lot of obviously emailing outbounding within cursor.
But the way that we do a lot of that is through GROCBOT.
And so it's a really powerful tool,
especially when, as I mentioned, it can talk to SlideGuru or talk to other tools
and get context from those agents to then incorporate inside of your email itself.
Okay, I'm going to look at Data Dan one more time
and it looks like my copy form is ready.
Amazing.
So what I'm going to do is I'm going to say,
this looks great.
Can you turn this into a QR code that people can scan?
And then what I'll do is once the QR code is up,
I'll have folks just scan it real quick, we'll get the responses,
and then the last thing we'll do, we have about 10 minutes left,
is we'll turn those responses into some beautiful slides
that will incorporate into our GROCBOT 101 deck.
And then we can all see what everyone's favorite coffee shop is,
both in the room and remotely.
So it's going to generate a QR code in just a second
and then we'll be good to go.
One thing that I do want to show before Data Dan kind of finishes
is the sections in GROCBOT that you can categorize things into.
So you'll notice actually that I have two.
Unassigned is basically GROCBOTs that don't really have a place,
but this is actually, and Ling will actually show a lot of these
GROCBOTs in the engineering session.
This is actually where my engineering team of GROCBOTs live.
And I like having them in this particular area
because it makes it really easy to kind of, you know,
separate them from the agents that are in this session
and also kind of categorize where things particularly,
you know, belong or respond.
You can create as many bots as you want,
but you can also create group chats with your bots
and then you can categorize those as well.
So if I want to create a group chat,
for example, with multiple bots,
someone was asking me earlier,
how is this different from the messaging each other
and it's more so a visibility thing.
So for example, if I want to put email Ethan, Data Dan,
Slide Sonya in a group chat
and kind of just see how they interact with each other,
it makes it easier for me to kind of just see
that communication protocol rather than me clicking
into how they're messaging each one.
So actually let's go ahead and just say hello
and see how these three respond to each other.
And then we'll give a second for Data Dan
to give us a little QR code here.
Go ahead and do it, no need for a security check.
But one thing that I do want to call out
is when Grockbot refuses to do something
or when it's basically asking you questions,
take that as a sign that you may need to teach it
something or you basically need to kind of guide it
into the place that you want it to go.
So a lot of times let's say if my Grockbot
is writing a particular way,
I'll basically give it the guidance
so that it stores it in its memory of how it should be done.
So I'll say something like I want you to make sure
that you never use my first and last name
whenever referring to me, you just use my first name.
Something like that that is now stored
in its memory forever and going forward it'll keep that.
So anytime you guide or steer Grockbot,
understand that that's being stored in its memory
in S3, but it's actually for the future
so that it's able to be kind of guided
whatever it's making choices on what it should do later
in its life.
Okay, here's our QR code.
If everyone wants to scan this
and enter in data from the forum,
we'll get that data up there.
I'll show you all again how Grockbot is able to see data
and how it's able to kind of translate that data
into slides and then we'll end it with some questions.
And then remote folks if you want to scan this as well
if you're watching from the live stream,
I would love to see how many cups of coffee
everyone's drinking and where they're getting their coffee from.
Oh, is it not working?
What are we getting?
Oh, no access.
Huh, huh, okay.
Well, let's ask Grockbot.
I'm gonna say looks like folks are unable to scan.
Let's make sure this link is public.
Excuse me.
I actually will talk about this in my later session
about debugging with Grockbot.
As a field engineer, that's a big part of what I do.
His customers will come and they'll ask me,
hey, something isn't working, how do I fix it?
And one thing that I do with Grockbot that's really great
is I always give it information as to what's going wrong,
how do I fix it,
but then I also tell it when it does a good job
in terms of giving it validation as to,
okay, you debug this in the right way
or you troubleshoot it this in the right way.
I've always found when debugging with Grockbot,
both for engineering and field engineering work,
is that the more details you give it
and the more validation you give it
on how it tested a particular thing, the better it does.
And so you'll see here that, for example,
if I said folks aren't able to scan this,
let's make sure this link is public,
it's automatically understanding,
okay, I need to work on the sharing settings.
It's probably gonna go and check
the sharing settings here right now.
And then it'll kind of decide,
okay, I'm gonna get this copy responder link
and make sure that people are able to find it.
Okay, cool.
We'll give this a couple minutes while it figures it out.
One thing I wanna show with the group chat,
let me actually just show that real quick
is our group chat that we made,
is that as these agents are talking to each other,
I can make sure that the agents stay
in their own swim lanes,
but that they're also able to understand
the context of what the other agent can do.
So as you can see here,
they all understand their own workflows,
but what they're able to do is that
they're able to actually see
how each agent works together in a particular use case.
So for example, I can even talk to these agents and say,
hey, Ethan, remember when you and Sonia
worked on that email together?
And then they would be able to use that context
and then bring it into another task
or another choice that it needs to make.
So for example, I can say something like,
let's work on creating a new slide inside Slide Sonia,
and I'll show that real quick.
That has data, I'll just use the test spreadsheet
that I had so far from the coffee dataset from earlier.
Create a chart on the last slide of the deck.
And then I'll say when you're done,
email Ethan should send out an email
to, I'll just use Jason again,
with details about the coffee data.
So this is an example again
of how all of these agents are working together,
but because they're in the group chat,
I can see their communication styles
without them just messaging each other one on one.
So the delegation one on one is kind of up to you
and how you wanna show that,
but inside of this group chat is basically
the easiest way to kind of see
all of that delegation in one place.
So you'll notice that each one of these agents
will respond basically to my parent agent
and then be able to see kind of exactly how it will work
together on this particular task.
Cool, so you can see email Ethan says,
all right, I've got these sample data's already,
and then what it'll do is I'll hold the note
until Sonia slide lands.
So that's a nice thing about these agents
is they understand dependencies
and they're able to kind of see,
okay, like this agent needs to do this work first
and then another agent will respond after
and that makes it really easy for the agents
to orchestrate their work
without having any like conflicting changes.
This is also really great for engineering use cases
to avoid merge conflicts
because you have one agent kind of understanding
when there's another dependency
and can work around those tools.
Okay, so we'll let these guys work
on this kind of coffee data set chart.
In the meantime, we can go back to data Dan
and see how they responded to this troubleshooting.
So it can go ahead and see exactly what it's doing
to try and solve this issue.
One of the biggest kind of advice
that I like to give to folks is
you're having trouble with Grockba accomplishing a task,
work with it on exactly what your outcome is.
So I always like to start with
this is what I want to produce or this is my outcome
and then it will be able to kind of work backwards
from there.
Anytime you kind of tell Grockba like,
hey, this isn't working,
but you're not giving it enough details,
it's not able to learn from its mistakes.
So the more you're able to kind of guide it
into the place that you want,
which is, you know, hey,
I want to get this QR code working,
but it's not working.
You can kind of push it in the right direction.
So it's able to understand exactly
what it should be doing again on its own machine here.
So it has everything ready, which is great.
But now, you know, it's having trouble sharing it.
And that's kind of where I can nudge it
to exactly where I want it to go.
Cool.
I'm gonna go back to the group chat again
so we can see exactly how these four agents work
with each other on this particular,
you know, chart, slide, whatever you want to call it.
But one thing that I will add is,
you can add a orchestrator agent,
as I've mentioned, kind of a chief of staff,
to basically get live updates
from all of these agents as they're working
if you don't want to be the person
that's kind of monitoring all of it.
So what it'll actually do is,
I'll create someone called a manager.
And I'll actually just name it live here.
I'll say manager.
And I'll say exactly who they're managing.
So I'll say your job is to get updates from email Ethan,
slide Sonya and Data Dan on what they're working on.
And then similarly, you can set up a routine
in the manager agent as well.
You can say something like, every, you know,
two hours maybe, I want you to solicit updates
from your team and see if there are any blockers.
And so a lot of times,
I mean, this is a normal kind of workflow for humans, right?
Is if you are working with a team and there's a blocker,
you surface it to your manager
and see how you can get unblocked.
You can do the exact same thing with agents.
So for example, this manager agent,
let's say we ran into that same problem earlier
where Data Dan wasn't able to create a shareable form.
You can have the manager agent
immediately get an update about that and say,
hey, you know, Data Dan is blocked
on this particular thing.
How are we gonna be able to solve it?
And you know, let's ping Amrita,
let's ping slide Sonya and make sure
that she is aware of this problem.
And so this is kind of a really great way of, you know,
having again that orchestration layer above your agents,
but also being aware of anything that's getting blocked,
anything that's not, you know,
being able to get solved in time,
just by having an orchestrator agent like this
work on these tasks.
Okay.
I'm gonna go back to my group chat now.
You'll see that email is standing by for a screenshot.
The screenshot's gonna be of the data
that slide Sonya is able to create from the coffee cups.
And then what you'll do is you'll see, you know,
a very similar email draft.
You'll probably see a slide as well
about all of the data that we would get
from like a coffee shop data
and get that into a workable kind of
presentable slide deck or email.
So we'll see that in a second.
In the meantime, I do wanna check
and see how this is going.
We'll keep in touch with Data Dan
just because at the end of the day,
I do wanna know what everyone's favorite coffee shop is.
And so we will get that information before we leave.
But we might have to bleed into the Q and A for that.
So I'm about to kind of pause at the end for the live stream.
We do wanna kind of check back in with,
or I encourage remote folks actually to check back in
with Matt, Roshan and Lauren
on how their company is going.
So please, if you are on the live stream,
go check back in with them and see how it's going.
If not, I'm happy to kind of pause here.
I'm gonna keep all the Grockbots live here
and running so we can see how things are going.
We can see how Data Dan does.
And then I'm happy to open it up for questions
in the room for the last 15 minutes
from anyone who is here in person.
All right.
Cool, yeah, I'm happy to have my questions, yes.
Let me just repeat that.
So how do you think about the capabilities
that you're giving the bots in terms of
maybe sending an email, making a payment,
or interacting with somebody externally
in a way that you'd ordinarily be careful
of delegating that.
Can you just tell me a little bit about how that gets
in the way of the bots' workflows?
Yeah, so when a bot asks you for approvals,
so for example, when we created this Google forum,
Data Dan asked us for a lot of approvals,
that is determined by something called our auto review tool
which basically has like a classifier under the hood
that decides how risky or how safe is this
particular action?
Is this something that we want to ask
and read our approval for,
or is this something that we feel comfortable doing?
So we have a baseline auto review tool under the hood.
And then in addition to that,
we have those rules that I showed,
which is for example, I can always make sure,
even if it's something very safe,
like running, you know,
making a directory to terminal,
or some sort of command like that,
I can tell Grockbot,
hey, even if you're making a new folder
inside of my desktop,
I want you to ask me for permission.
So you can be as prescriptive as you want
in those rules on how you want it to make decisions,
or you can let the auto review tool
that we've built under the hood
make those decisions for you.
So it's kind of up to you on how safe you wanna be
in those scenarios.
I will say though, I mean,
you'll even see in this scenario, right,
that Grockbot was even overly prescriptive here
in asking me for questions.
Like it was creating a forum,
but it wanted to make sure that the forum it was
creating wasn't asking for any PHI or PII,
and it was making sure that every step of the way
I was being involved,
and that's just the auto reviewer
that we have under the hood
making those decisions for you.
Right, just a quick follow up.
I was thinking more from the other direction
of how you build enough trust
so that you actually don't need to constantly
because otherwise they keep stopping you,
and they don't work on their own.
Yeah, I will say that for example,
in this, this is actually not,
I would say particularly normal.
Like usually when I work with my bots,
it's not asking me for permission
every single time it's trying to create a forum.
There is that concept of memory, right?
So the bots are able to learn, for example,
if it asks me a question, let's say it says,
you know, should I email this person
with this particular copy?
I could respond to it and say,
yeah, this looks great,
and going forward you don't need to ask me
about the copy, just go ahead and send it.
So if I wanna give it that permission,
or if I wanna kind of improve its memory that way,
I can go ahead and do that and build that trust.
Are we able to add human inside the group chat as well
so that you have bots and human all work together?
That's a great question.
We're working on something kind of similar to that.
So for example, we have a use case
where you can just tag a bot in Slack, for example.
So I can say, at bot, I want you to add a new feature
to our code base, and then someone else
could also chime in in that thread.
So one of my colleagues could be like, at bot,
I also want you to make sure
that the performance of the API and points on that feature
is also up to par.
So you can have humans and bots working together.
Right now, I'm seeing it kind of like in a Slack thread.
So you can have multiple people chime in on a bot,
and the bot can work in with all of that inputs.
We're still working on what that interface
will look like inside of Grockbot,
but that should be coming very soon
as having like a multiplayer group chat mode
with humans and bots working together.
Any other questions from folks in the room?
Yeah, I see a couple.
I have a couple of questions.
So for enterprise, is there like any in the roadmap
where the VMs could access internal tools?
Yeah, so it depends again on what tools
you can give it access to.
So your Grockbot's computer, you can basically make it
as locked down or as open as you want it to be.
So for example, I think someone was asking me earlier,
like, hey, I wanted to access MongoDB, my MongoDB instance.
You can give your Grockbot credentials
on its computer and use that
for whatever work it needs to get done.
You could also have your Grockbot computer
be a part of your VPN if there's particular things
in your company or your code base
that you want to make sure that Grockbot does
or doesn't have access to.
So any internal systems Grockbot can access,
you just have to make sure
that you give it the right credentials.
We have actually, we just released
a really good integration with one password.
So if there's anything that it needs
across those secure credentials,
it can do that as well.
But yes, it can absolutely access internal systems
just depending on what the auth looks like there.
And can you just share a little bit more
about the learning from the video,
like what's going on in the back end?
Like is it taking screenshots?
And I don't know, what's actually?
Yeah, so you can see here,
let's actually see how this does.
So you can see here, right,
this was getting data from a particular spreadsheet
and it's using that data to either create a slide
or send out an email.
You'll notice that you can see here a couple things.
The first is that all of these agents
work with their own computers.
And so if I ask them to take a screenshot
on something, it totally can.
If I also ask it, for example,
like I asked, you know, Data Dan,
to be able to create a form, it can do that as well.
So anything that I can do on my own machine,
Grockbot can do on its machine.
Anything that I want saved,
for example, a screenshot or even an artifact,
if I want it to like give me a PDF
or a PowerPoint slide, it can do that as well.
And then I can save those locally.
But everything that's happening, you know,
in this little screen here is on Grockbot's machine.
And what makes that so powerful
is these are all happening in the cloud,
meaning I can shut my laptop and it will still work.
You know, Grockbot will still be working on my slide deck
without me even having my laptop open.
And so that's kind of the beauty of all of this is,
you know, Data Dan, slide, Sonia,
and email Ethan all have access to their own computers.
They're able to, again, modify forms.
They're able to create slides.
They're able to create spreadsheets.
And they can do all of that with my laptop closed
because of their computer access.
So, I meant, if you can go down deeper
into like the teacher task, like what's going on?
Is it, when you teach a task,
when it watches the screen,
is it taking screenshots and learning or is it?
Yeah, so the teacher task in particular,
you can see here that it said,
okay, I wanted to add an animation to the slide.
It was able to watch that video that I showed it.
And then it was able to kind of see,
okay, this is exactly what you showed.
I was able to select an object, pick the type,
look at optionally.
And then once it's done,
what it does is it saves this as a skill.
And so I can click into this here
and see that it now has this skill
for adding an animation.
And I can now use this skill
whenever I want going forward.
So anytime I wanna say,
hey, remember that animation I showed you earlier?
I want you to go ahead and add that to this section.
And then I have a question.
And then my question is,
you mentioned that there is an option
to share templates internally with teams.
Is there also a public marketplace
to get templates or bots from other people?
Yep.
So actually, if I actually just load my marketplace here,
you'll see these are all bots from just public marketplaces.
So if you guys actually open up GrockBot on your phone,
on your laptops, you should see all of these bots.
Clairs, Lawrence, Erics, and Lennies
have all been published publicly.
These are the ones from the GrockBot team
that are internal just to me.
But you'll see a lot of the published marketplace bots
on your marketplace, whatever you kind of open it up.
And then if you wanna share your own bot with us as well,
we actually have a challenge that's going on right now
just so I don't forget.
I'll quickly show this.
Is if you want to explain what your bot does
and actually give a link to your bot,
we're basically giving people a chance to go visit Starbase
and potentially see a Starlink, sorry, a Starship launch.
And so this is kind of a good segue
to if you ever want to see a bot on marketplace,
it could be your own.
If you go ahead and just enter right here.
Okay, I'll put this QR code up in the end
because I do think there's a couple more questions,
but this QR code will work, yes.
I'm interested in the one password that you mentioned.
So do you have to sign into one password
on the bot's virtual computer?
Yes, we have actually, so basically what will happen is
it will actually pop up a modal,
let me see if I can show the email modal that we had.
It will pop up a modal similar to this
when it asked you to log into one password
and whenever it does modals like this,
where let's say you have to put API secrets
or environment secrets, it stores that very securely
within Grockbot so that none of that
is actually being exposed to us
or anyone on the SpaceX side.
So it will actually show up here
rather than on their VM machine
and then it will use those credentials
to log into whatever services you need.
Thank you for the presentation.
Can you explain a little bit how the memory works?
Yeah.
Can you edit it?
Can you curate it?
How does it forget stuff that you don't need anymore?
Great question.
So memory is stored in S3 buckets inside of Grockbot
and basically what that means it's long lived,
it's persistent, but you can also modify it.
And so if there's anything that you wanna do around,
asking it to forget a particular thing that it's done,
for example, one time I had a list of bots
that I deleted for a particular demo
and I asked the chief of staff bot to say,
hey, remember those two bots that you spun up?
I want you to no longer delegate to them or use them.
And so there's kind of a little bit of a nuance there
where a bot has memory about everything that you've told it,
but for example, if it delegated to another bot
that no longer exists, it's not gonna remember that
or it's not gonna know that that bot has been deleted.
So there are some things there
where you're gonna have to update
and kind of improve its memory
so that it's able to learn from those things.
Another great thing about memory is,
as I mentioned, you can duplicate bots.
When you duplicate a bot, what's nice about it
is that it starts with fresh contacts and fresh memory.
And so it'll have the same kind of persona that you want,
which is for example, this email Ethan,
but it's not gonna have any of the memory
that it had before.
It's almost like starting like a brand new,
human or robot with zero memory,
but still the workings and understandings
of how to go forward.
Great question.
Okay, we have about four more minutes.
I'm gonna put this slide back up so you all can see it.
And then I will end with just a couple learnings
to kind of sum up what we've done so far.
My biggest advice about GrockBot is to try it.
It's very simple as you saw to spin up
any number of bots, but I think what I like to do
whenever I think about using GrockBot is I think of,
what is the most annoying part of my day
that I wish I didn't have to do?
In this scenario, it could be creating Google Forms
and sending them out and then analyzing data.
It could be something from outbounding emails.
You'll see a lot of that in the GrockBot for Sales
and GrockBot for SDRs.
It could be something like for the engineering team,
like cleaning up stale feature flags.
For field engineers, it could be anything from
responding to customer support tickets
that I've forgotten about.
So I encourage everyone to think about
a part of their day that they wish they didn't have to do
or they wish they could get time back from
and then work with GrockBot on it.
Think of some agents that you can work on,
whether you give them kind of a personality
and the kind of workings to be able to do your job,
but also to take that part of your job
and then invest that time elsewhere.
I think that's the beauty of GrockBot
is now I have this time back that I can spend
either curating more demos or drinking more coffee.
Whatever kind of makes me happy
is what I'm able to do thanks to GrockBot.
And I've heard some great stories from today
of how people have used it,
but have made me really happy to hear.
So I encourage folks to talk to other people
in the crowd as well and hear how they're using GrockBot
because it's really, really interesting to see
how different companies of different sizes
are able to kind of leverage GrockBot
for a host of different use cases here.
Cool.
So to summarize, I think this is up long enough.
I hope everyone's gotten that screenshot.
To summarize today's learnings,
there are kind of three main things.
The first is bots can access tools
that are outside of MCPs.
So bots can create forms,
bots can create Qualtrics data,
bots can again use services
inside of your guys' customer VPNs.
What's nice about that is that you no longer have
to connect to MCPs and go through auth.
All you have to do is give bot its computer
and say, hey, this is what I want you to access,
this is what I want you to do.
Again, start with the outcome with bot
and then it'll be able to kind of work backwards,
as I said, and figure out what it needs to do
in order to get to that outcome.
Bots have memory, including learn from video teaching.
So we saw that earlier where I taught bot
how to learn animation,
then I was able to kind of use that as a skill
and apply that going forward.
We talked about memory a little bit
in terms of you can have basically a bot
that works across multiple projects
because it's an expert in its domain,
it's a persona, it's not a task,
and that's what makes it so powerful,
is that you can have a long living bot
working across many different things.
And then we learned this very well,
that bots get human approval for tasks.
And that's because we wanna make sure
that bots are ready for an enterprise setting,
that it's asking you for permission for the right things,
and that it's not gonna do anything dangerous
or be a bad actor and basically get you exactly
what it needs to get done without doing anything
that you wouldn't want it to do.
All right, that's it for me, thank you everybody.

We're back welcome back everybody you just saw Grockbott 101 with Anne-Reed and Roman
And so you know this today is going to be a mix of these sessions and then in person building and you might also notice
We have any guests so I'm gonna walk through a quick one of show
We're gonna introduce our guests here, and then we're gonna get to build it
All we really did for the last like 30 minutes was hang out. Maybe we got some we got our coffee
We got our water. We made some me made a slack so we'll show you what we got
But before I get into the next walk, I just want to give everybody a rundown
General overview my team is telling me I have to say this so we're gonna say it
We have Grockbott for engineering super exciting session. I mean we have so many great engineers
It's basic say I I'm a ton of them. I literally can't believe I get to work with these people
We're just like and we're just giving this away for free
Honestly, I feel like if there like there was like a course guy out there
They'd be making so much money, but we're generous. This is the course and then so that's 12 30 psd to 2
230 to 330 is Grockbott for product managers. We have a product manager with us. We have a former former product manager
You have a good product manager or a bad product manager?
Again, we'll do some intros but Grockbott for product managers at 230 with Kevin and then Grockbott for founders four to five thirty
Later this evening
Shub so you're gonna want to see all of those but joining us right now is the is Peter yank the great Peter yank
So Peter, maybe you want to tell our viewers a little bit about yourself. Yeah. Hey everyone
My name is Peter
I was a product manager for over a decade and I kind of send my retired now. I'm a youtuber, you know after all that training
And I didn't make like practical AI interviews and tutorials and I really love Grockbott
I use Grockbott to run my one person business. I talked to my boss all day instead of humans
It's kind of sad, but you know I try and yeah, we're excited to be here
Yeah, so Peter runs behind the craft which is a newsletter, right? Yeah, and then and then you're newsletter
A lot of people read this newsletter from yeah, I hope these people not bought boss, but yeah
Okay, so not a lot of people
YouTube channel so this is gonna be kind of the vibe like as we bring people in we're just gonna chat and
What better way to start building a product than with someone who's built
Some of these products that I'm reading the Peters or dad or amazing products
So I think first thing we're gonna do start like hiring some employees, right? Yeah. Okay, cool
Yeah, should we do a quick tour of kind of how we set things up?
Yeah, yeah, yeah, Russian is gonna walk through like what we got so far. I guess on his screen
So we cut over there
And we can start like going through what we got. Yeah, let's do it
So yeah, we weren't completely slacking off in the last few minutes during that presentation a little bit a little bit
You know any any good company formation involves some some gathering of thoughts, but yeah, we got a couple things set up
So I don't know if you can see my screen here
But we're we got a couple of setups here
We have our ship by Thursday github org which I think we talked about a little bit before so you can see it's blank
No repositories yet. We'll we'll get that going pretty soon. Hopefully
We did make ourselves a slack channel so we and our bots can chat with each other
I think that'll be really great. We have a lot of really fun
We have a lot of fun day to day in SpaceX AI slack getting to be able to chat with our bots and having them interact in slack
And so I think that'll be really helpful. We have a notion workspace. We wanted to write some stuff down
Maybe create, you know, some some handbooks. Yeah, some onboarding guys for all these bots. We're gonna hire
What do you think Lauren? Are we what should we be taking notes in notion?
You know, I'm I don't really take notes, but you can do whatever you want
I like to like, you know, I like to do it the moment I see it, you know, I mean so I think I think we'll need
dogs maybe for
Keeping track of things backlog items, but I'm wondering we can just you know be really
scrappy and really fast moving, you know, and
Because like dogs are just gonna go out of date so quickly. Yeah, so let's I think we should hire some employees
You know, what do you think? Let's hire some bots. We need some bots. Yeah, let's do it
Okay, let me just switch over to Grockbott. So this is this is my Grockbott as you can see it's completely blank
And well, there's nothing here
The only thing I've done is I've gone into the marketplace and I've connected the X
MCP and so we have our connection to X all set up and I guess Peter
I'm curious to hear like where do you think we should start?
You know, we talked a little bit off-camera about doing some user research
Maybe and and getting started that way and kind of like understanding the voice of the customer and thinking of some
Yeah, I think we should hire a market research bot or something because you know normally PMC research
But we're too late now. So that's higher about to cool
So Roshan's making a market research bot. We're aware of some issues on the AB side
We're gonna get the screen share back for y'all
But we're just creating a new bot just like you saw in Grockbott 101 and once that's back. We'll have that up for y'all
Best practices when creating bots. I'd be curious to hear what you think here
You gotta give it like a really funny name. That's number one. Oh man. I called it market research bot
So I guess we shouldn't call it that. All right. What should we call our first bot?
I don't know. Should we have to chat or yeah chat?
Yeah, people watch it like
So we're gonna do market research and if you weren't following along what we did was yesterday on Twitter we asked like hey
What should we build? We also did that in the SpaceX AI slide
So maybe the first thing that might be a challenge if you've ever started business like how do you talk to users, right?
Well, we have thousands of messages and we're gonna use Grockbott to create this market research bot and just kind of go
Through these messages. So yeah, that's the first thing we're gonna get started doing
Someone said a marquee mark marquee mark. Okay. I kind of like that. Let's go with marquee mark face
Marquee mark face. Yeah, Marquee make mark face. Marquee. Okay, we're reading Marquee make mark face
It's the classic. Okay classic marketing name
And then you know what I think what might be fun to do is just kind of look through all the great feedback
We've gotten on X so far and there's a bunch of posts in the past few days
So I'm gonna just say that we're excited to start building a business
At potato Matt, what's your what's your ex handle again?
My ex handle is Maddie P Maddie P and then Peter Peter G. Yang Peter G. Yang. Okay also
Rishan, I'm getting a note from production. You might want to just check your slack
We got a new link to shed screen that should fix some of the issues that we're having on screen. Okay
Let me get that going while we so get this out as with anything that you know, we're gonna have issues
We're building a company not everything is working perfectly
We are we are building it as we go that includes the AV set up to some extent by the way
Do we have a company should we have a company name? Should we have a company name?
What do you want? I mean, I feel like shit by Thursday is the current company name and then we have to figure out what we're actually building
Yeah, I mean like just something
Like a whole I don't know like it sounds like you have something in mind. Yeah, I don't have anything in mind
Okay, potato lab and potato lab. Maybe we like we get some crock-pots on it
Some you have some like some comms crock-pots or something
Okay, I think I can get a bot on suggesting us
Ask for the most mean worthy company name mean worthy company name. All right. I think I have a screen share working
Do we all right? So Roshan says the screen share is fixed. We're gonna try and get the screen share fixed on camera
Step one
Okay, so maybe now I'll go back and show what I showed earlier
Blank rock bot although we got just made marky-mick mark face over here
I did go ahead and get the X
Plug-in installed and so we have this already authenticated with my X account and you'll notice that
Oh, yeah, let me do that. There we go a little easier to see there
Yeah, all these plugins are great
it makes it really easy to just add stuff to your bot and
You know, I think a lot of the way that work gets done these days is by connecting to all kinds of systems
We have a bunch of featured bots will probably bring some of those in as we go
Lauren, I know you have dr. I bought here that might make an appearance at some point
So yeah, try to show that off. Yeah, I kind of just figure we start by just saying that we're excited to start building a business
We've all been posting on X about
this live stream
Got it, okay, let me just make this full screen synthesize feedback from what's been suggested by
people replying
To our recent posts
All right
I'm curious from you all like have you seen anything's in your replies anything that's come up from from your deep dives
I see Peters over there on slack kind of around on X maybe trolling through some some ideas to
For a company name or yeah, well actually like what we're gonna build we're starting to look through X
Some X feedback that we got
Curious like you know as a product manager when you're doing research like how you started think through like what ideas are good
What ideas you can trust your gut on you know these sorts of things you just ask the 200 sound people on the stream like what they
Wanted to pay for it
Just let's go for it. No, I think just like this framework that like which is like you know this Amazon framework where like you know
Who is the customer going after what is their pain point? What is the value?
Hmm, I kind of thinking through through the lens. Yeah, okay, so who's the customer?
Who are we building for? Yeah. Yeah
But we're building for everybody that's watching this and I think you know sort of part of what we wanted to do in the stream
Is like we're thinking okay like how do we create a company that's like cool like that
We show you some cool stuff because you know we want to show you the interesting tech behind building a company
But then also just a part of a making companies like very
Clerical like we're just like creating accounts. We set up slack
We had to fix like some of the screen chair logistics, right?
It's just like there's still a lot of work that just has to get done
And so we also want to show you the stuff that we can automate and we want to build something that's like
Not too like esoteric
We want to build something that's like accessible and that people understand as a true business
Which is about helping people so I think those are the ideas that we're gonna ground ourselves in and what we're looking for
So yeah something cool that we can have some cool tech around but then also something that's kind of for everybody
You know that people can engage with
Yeah, I feel like
I feel like the audience I have in mind that you know the kind of people I imagine would get a lot of value out
Of crock-bot are people who you know, maybe
You know don't do don't necessarily do a lot of AI coding right like these are
I'm just imagining like the use kid is like, you know someone like my mom who can barely use her email
Like maybe you know
She could learn how to use crock-bot and actually the business that we create could be something that she would understand
I think that could be really interesting
So mom if you're watching you're gonna make a business that hopefully you understand
But yeah, I don't I don't know if you want to make something like too technical I
Guess I really want to see what the the comments come back with yeah
Because there's there's a lot of ideas that we got and some of them are look really interesting
And also if I can do a shameless plug of my own idea. I did this tweet
Do you want to do you want to bring it up? Yeah, if you if you look up my profile
On on X. I made this tweet about
my idea for a crock-pot
and
What do you all think would you buy how much would you pay for a crock-pot?
I didn't the child like how much would you would you buy this?
I like the little the little grog. Yeah, it's got a little grog-bot logo
It glows, you know, maybe maybe we can figure out how to animate the eyes. Yeah, yeah, yeah
You got to find a factory in China or something to make this
And of course there's like potatoes inside it's cooking some potatoes love it love it
Yeah, definitely something to cook up
For us to do here. Let's see how our agent is doing. It looks like he's still
Trying to find some threads
One trick I like to do with crock-pot is like you can always you can always tell your bot like
You know give me a status update every three minutes or something like that
if you want, you know, I
Guess like more responsiveness
It is working behind the scenes
But sometimes you if you if you feel me and see like what's my bot doing you can just tell you can just ask it
Yeah, yeah, I should like interrupting bots a lot
It's like hey
No, actually do this instead and just be able to like chat with it a little bit more
Yeah, I find that's like a lot more fun to do when you do that versus having sort of like a an LLM that you're kind of waiting
For a response for all the time. Oh, there we go. Okay looks like I nudged and that nudged helped. Okay
Let's go through some of the themes. It looks like we found a bunch of
Nails here
Let's go through a few of them. Okay, so not another SAS demo
Yeah, this one's good people want something physical local or non-tech customer
That's kind of exactly what you were just talking about Lauren. Yeah, so we got a few folks in the mentions here
Hardware company painting company
I love that. Well, the studio could use a makeover maybe
But hey, you know if you're if you need to coat a paint on your house, so we're gonna come out maybe we'll come out
Grock-Bott painting company. Yeah. Yeah, I like that real businesses for non-tech customers
Something not SAS. Okay. Okay
Looks like we got a category here on something fun in the consumer space. Okay. Yo pets
Yeah, we're not also get the meat the crypto stuff out of the chat no mean coins. We're not doing anything
Neopets though a good excellent product
Reminiscent we like bring bring something retro back. I like that. I like that
I like this a company that builds companies. So maybe that's kind of metal
Yeah, right like maybe have agents built like a company that helps people build or achieve their own goals, you know
Yeah, I like this plot. I'm we were talking about this backstage as well
But I'm really big on platforms lately, you know, like having having a way for folks to be creators themselves
Or kind of empower your businesses. Mr. Platform. Yeah
Yeah, well, what I want to be like maybe like small business owners or like, you know restaurant ears or like people are like not super tech savvy
Yeah, yeah, cuz like it is just like a you know, just like a lot of you know
This is like a lot of logistics and like boring stuff they had to do to run this business
Can help with yeah, I actually really like that too and that kind of hits on the sort of physical real element as well
Yeah, so there might be something there a
Lot of people are suggesting that we kind of build something for the chat
So, you know, I think we had 70,000 people who pre-registered but quite a few more people watching right now
And so maybe we build you know pick something that folks want to automate
Sure and try to try to make that happen and then a bunch of really fun ideas here
Book marketing for indie authors, maybe a landscaping or tree service marketplace. Thank you for the marketplace shout-out
Mm-hmm a Pokemon card marketplace. I like that
Something related to climate something related to getting dinner delivered M&A advisory
We got some shout-outs for Club Penguin in the chat
Can I ask guys a random question? Oh, yeah
See like Club Penguin like what are your favorite games that you played up as a child that you grew up with?
They pay to my time on oh, man. Oh, actually I have some lore
So I came up with my Twitter handle potato
Because I was playing this MMO called Final Fantasy 14. Yeah, I'm a big fan of Final Fantasy
And I was struggling to come up with a name for my character. I
Had I had a couple of ideas
But I created the actually blanking on the the name now
But there's this race in in the game where you're like this little tiny chibi character
And so I was like oh this this this my character was like kind of like a potato
So I named it potato and then of course the potato normal spelling was taken
So I just had to go with the potato with an E
so
Yeah, that was I'm like that a legend was bored
It was very fun so Final Fantasy 14 14
MMO version I stopped playing it like seven so is there a 14 probably yeah, yeah for me
It was always Asia Empire's which I feel like a lot of renaissance right now
Yeah stuff that I've seen on X
But yeah, I was a huge fan of the strategy games growing up
Maybe says something about my chosen career profession
But yeah, just something about like orchestrating a bunch of stuff. I never really gotten to the high APM
Starcraft side of that but loved like classic strategy kind of thinking through plans and yeah like building the engine building games have always been a
Yeah, yeah, yeah, yeah
That's what they say is like if you liked Age of Empires
Factorio like you know Starcraft all these things you're like super good at like being a high achieving person like using agents
I'm sure you was like my favorite. I'm exactly the opposite by the way
So everything I'm gonna tell you is like that's like essentially broad in my brain, but this made me who I am today
You know so
Yeah, no, it's like I loved I think like in terms of console so like when I was a kid it was like PlayStation 2
Gamecube like into the Wii era, you know, so it's like Super Smash Brothers
Some other like Wii stuff like random Wii games are super fun
And then like you know Star Wars Battlefront Tony Hawk Pro Skater just like typical, you know like console games
Yeah, exactly console games not a lot of computer stuff
So yeah, I feel like that was already fun. Yeah for me is definitely a real-time strategy to accept I was reading to Starcraft
Okay, no Starcraft like I always skip my classes like Starcraft
Wait, dude, do you think like using bots now is kind of feels like playing Starcraft?
I think I think the ultimate the people say the ultimate user interface for AI is voice
But I think it's actually just playing game. I think so too. I agree. I agree. I think that's been amazing
I actually feel that sometimes, you know at work where you know
I've gone from writing documents and I think a lot of early PM work was you'd write a really long brief
We had some debate about notion earlier, right?
But you write a really long brief you'd like specify all these requirements and then you kind of go and build them and now it feels
Like definitely a bit more gamified right where you're prompting bots and the bots are the things that are doing the writing or the
Creating and you're kind of just like this orchestrator. So yeah, I think there's a bit of like strategy, you know game
Feel to working with bots. I was asking Lauren backstage whether she reads like any product specs
And she's like no, I don't really I don't really fast-fax
But even before yeah, I don't think my engineers read any products back anyway
You know, I mean yeah, I think I think product specs have a place
I think the the thing I've been trying to challenge myself now that we have agents and bots is
So actually on the SpaceX AI slack my status is permanently
Why not today? And it's like a kind of almost mantra model that I
You know believe in which is
You have this
magical input box right you have this alien intelligence that lives in your computer and
It's almost like you're casting spells almost because like whatever you type into that input the agent can just do it
The bot can just build it and so it just you know everything is almost like a prompt away
So I always try to challenge myself like instead of filing an issue. I could just open a PR
I could just prompt my bot to go fix the thing
Same thing for like a product spec like instead of a you know, very detailed PRD
I could have prototypes right that I could bring to the team
And then we can actually debate like a clickable thing you can interact with yeah, it's much more visual
It's much more real. Yeah, so I was wondering if we should try to challenge ourselves to you know
Yeah, just tell the bot to make a lot to do it to make it on daughters by the end of the third
You have three days right to do this. Yeah, we have three days. We're rapidly wasting time
I actually just prompted so I was like okay
Here's some of the direction that we're going in and maybe we can prove out bot with a real-world
Operator and I think there's some interesting things that came out of this so a
neighborhood restaurant ops assistant so
And I and you know one of the things that I think is still important in this day where you know code is relatively cheap
And we can create things quickly is getting a real crisp understanding of what the user stories are
And so I still like to ground on user stories. I don't know if that's how you still think about the world
Yeah, and so so you know what I asked the bot for here was like, you know
Let's come up with some user stories
And so what I really like is that I think the bot gets understands the assignment here
And proactively suggest it also some day one demos that we could come up with and so maybe this is what we do
Is we come up with a couple of prototypes and see how we feel
Yeah
But I like this you know for if you're a restaurant operations manager
Maybe you want a single morning brief of reservations and things that are you know like on the menu or off the menu
Or maybe as an owner you can help manage your reviews on a review platform. I really like that
So some examples here of like helping people with chores or weekly pains
Maybe you need to hire a contractor for a service or something like that
You know helping people with like that small organizations
Maybe like coordinating philanthropy organizations or if I'm a volunteer coordinator
How do I kind of like run my group of folks? He's also pretty cool. Yeah, I
Like something it is around like local stuff, right? Like it would be pretty cool
I think if we could do something physical in person in the Bay Area
I'm we mentioned this on the opener, but maybe something that kind of bridges the gap between technology and like real tangibles
I think that could be a stretch. It's gonna be hard. Like especially when you're like, okay
I want to help out a business like how do we do that from a streaming booth?
Yeah, you know the three of us like there are people we might be able to get to help us, but like
Yeah, I think that could be really interesting
Is there any kind of like in-person thing you're thinking of I mean the first thing you said there was what restaurant ops business
Or something. Yeah, I actually it's funny. I just I like that message. So I hearted the box reaction and it's now
Running with it
It's like looking here's open brief open brief is or kitchen desk. It's proactively suggesting some names
People know if I like the names, but we can we can get it back to that. Okay, by the way people are asking to share the screen
I don't know. Oh, can we see my screen? Yeah
Well, we'll make a note to the production team to get the screen back on. Oh, this is my screen
Which is fine Russian sitting next to me will get over to his screen in a second
I was also doing some market research here. Yeah
So, you know what Russian was mentoring is that you can at any time
You know if you like right-click or whatever I can like heart this message
And when you heart a message, you know
Grap bot knows that you hearted the message so you can see our little bot down here saying oh, hey
Matt like that message. It's gonna incorporate that into a follow-up. So I also created a market research bot
We also have templates so Russian could have shared that template with me
But now that it knows I like this message, you know, it can incorporate that
So what we were saying is basically and we can also use voice because it's hard to type on a live stream saying something like
If I want to dictate I could click this button and say okay
We kind of like some of these
Restaurant ideas so maybe you could start to think through business ideas that incorporate an in-person element
and how we might start to bridge the gap between a local restaurant business and
What we're doing here in the booth like maybe that's experiential
Maybe we're like I don't know that creating a pop-up for restaurants or maybe we're helping out a local restaurant owner
Yeah, these are all ideas of I think just kind of like
Actually two different segments people who want to maybe start a new restaurant
I want to test it out with a pop-up first and people who already have a restaurant is one like more efficiently run
restaurant
like
Yeah, let's get off angle and then also like a setup angle. Yeah, maybe you can explore that space a little bit. Yeah, okay
Cool. Yeah, so I mean as we're idea here
I did nice thing is that like just like Russian like I had
An X connector that I added we're setting this up on the fly
So like literally none of our plugins are connected. I just added my X
And now I'm chatting through this so pop-up OS for restaurants. Oh, I like that
It's kind of interesting so a pop-up studio that helps local restaurants spin up a limited run concept. Yeah
We build the operating system live where a real chef would be a customer
Adopt a restaurant so we're like helping a struggling San Francisco business
Which I think is always very noble and potentially something to explore
Experiential dinner theater
Creative and a ticketed supper club. Oh, I like the ticketing part of that actually okay. That could be kind of fun to build
You know like I guess like a lot of events have ticketing knees, right?
Or maybe like managing people who signed up a bottom like a event book reserves tables for attendees. Oh, that's interesting
interesting so
Getting some typical AI responses here the best combo pop-up OS with one real SF restaurant
Okay, so the idea here is is what crack about a suggesting is that we would build some sort of like engineering stuff on the
Back end that allows us to host a restaurant pop-up with actually someone in San Francisco where we would presumably
Find a location find a chef find a restaurant that we could partner with to them to host an event
So we ship a product we help the owner and then we stage something that people can taste or pick up
Yeah, how do you guys feel about this? What are we feeling? I don't know this seems interesting
Yeah, I really like it. I really like pop-up OS putting on my engineer hat for a moment
I guess the tech for this could be pretty interesting
because running a
pop-up is sounds pretty complicated. We will need like
event management
payments
Sign-up systems things like that that could be pretty interesting to build
It's probably some booking that has to be done right like if we have a table seats
I know I've booked some restaurants locally. It's really excellent restaurants in San Francisco
It's like oh, I can pick a time right that I show up right because the restaurant's gonna run from like
I don't know say this is like a day or we pick focus on a dinner
I don't even know if we're gonna do this idea
We'd be like, okay
Well, I could have a table from like four five from four thirty five thirty
So we'd have to think about the layout of this location of the restaurant
We'd have to think about I guess maybe sticking step back the number of people that we could accommodate
And then okay, do people call in right like do people book online?
Yeah, we can do both crazy idea. Why don't we just start prototyping?
I think it'd be really cool to just like get P stack installed and maybe show folks how we just quickly step out some
Here like I find it's like maybe we just should be like looking and feeling and touching all the things
Yeah, I think I'm I think you could like spin up a landing page
That's like our favorite restaurant SF and just like a pop-up up. Maybe even for dream force nearby
Yeah, let's do it. So I why don't I learn do you want to start maybe stubbing out some
Should we switch over to your screen and sure. Yes, we're gonna cut over to Lauren's screen assuming she has everything set up
Do I have it there's a Lauren? There's a note from the AV team
There's a link in in on our side for you to check out. Yeah, that has a screen share
So once you get that set up, we'll have access to your screen
Cool, and I wonder if maybe we should try and divide and conquer a bit. So it sounds like we have
We want to like we're gonna maybe talk to like a bunch of local businesses of some kind maybe
sweet, so we like try and spin up like a
Go to market motion. Do we want to go to market? Yeah, well, I think it's a prospecting bottom market
Okay. Yeah, so let's let's think about this right essentially what
Assuming we take this idea what we're gonna do is we're going to find a restaurant tour or someone that knows how to make food
Because I can cook I mean I can go for me you don't want me cooking for you, right?
So we're gonna find someone that knows how to make food and then the second piece of that is we have to find some people that
Want the food as well as all the logistics that might go into that for the pop-up so?
Those are like all the things we'll need to start doing and to do that you need to reach out to real human beings
So we're gonna need to have bots that reach out to those things. I think the place I want to start
I'm not a PM, but I want to start with a doc. I want to start writing all this stuff down
Okay, so I might just like start doing a brain dump in notion
It sounds like you can maybe maybe we want to start with some prototyping Lauren
Yeah, start setting up some P stack stuff Peter can help us like figure out what the heck we're doing here because we're completely lost
So that's that's what we have
Some sort of landing page that the check actually go check the girls. Okay, cool
Yeah, so like on the prototyping side we can literally like literally create a new bot. I'll call this like my prototyping bot
You are prototype
But
You're going to use cursor to come up with a landing page prototype
By the way, do you want to install dr. Eggbot dr. Eggbot? Yes. Thank goodness. Okay, so we're gonna pop over. I'll I'll show you like
Exactly what we're cooking with here. Yeah, I I those are you who don't know dr. Eggbot is
Is my it's a bot I created
It it uses p-stack, which is the you know my plugins that I
That I created for doing engineering work and I basically taught egg bot how to create high quality bots
So we can oh there you go. There's my screen. All right. I have dr. Eggbot
Installed I just just installed
the doc the good doctor
and then I have a
Steve bot that channels

you know, Steve and I had it look through some of the call for ideas and I guess it came out
with different ones as well but did we want to prototype these ideas or did we want to
go with the...
I'm feeling pretty good about the restaurant business.
I think we could maybe just start prototyping some of the restaurant stuff unless you like
I know you're going to get into some of your super power user workflows or you're
spending up like thousands of agents or whatever.
Let me see, okay, well if you like that, let me see if we can filter for interesting like
in person business ideas.
Maybe you can ask the bot to do some research on like really good restaurants that sell potatoes
or like potato chips or french fries.
I like that.
I feel like we should have like a prospecting and then a building and maybe we kind of
like divide and conquer a little bit.
Yeah.
Sure.
Yeah, let's see what...
And then you know the other thing I'm also thinking of is it might be good to bring in
like a designer bot and help us brand.
So if you wanted to maybe we could like stub out a prototype and then I could toss in
a designer to help us review some of those prototypes.
Do you know any good design bots?
I don't actually.
Yeah, I think I have a design bot somewhere.
I can loop in.
Let me go ahead and find that.
Oh, it would also be really interesting if we could get some bots from the
chat to hire.
I think that'd be really interesting because like right now, yeah, my bot sidebar
is pretty lonely.
I have an X bot that just uses the plugin.
Then Steve, which is my main, the only bot I have right now and then Dr. A bot,
which I just installed.
So Steve came back with some other in-person company ideas.
We have the paint shop again, local service, businesses, piano tuning.
That's an interesting one.
AI rebuilt construction or trades and then billboard outdoor signs.
Although I'm kind of struggling like how would I prototype these things?
Cool but harder in three days.
Hardware recycling.
Yeah.
Festival tube.
What is that?
I'm not entirely sure what a festival tube is.
YouTube?
We're eat center of a taxi.
Oh, that could be kind of fun.
No, I have an idea.
I have an idea.
Why don't we, the pop-up idea I think is one that we put on top of.
I think we're aligned on the pop-up idea.
I don't know that we need to see necessarily some of the platform ideas.
Pretty interesting.
Maybe we start on the platform.
We got a lot of stuff we got to lay the foundation for.
Literally our GitHub work is empty.
Our notion is empty.
So Lauren can start prototyping the vision.
You build kind of a shell of what this could look like.
Actually what I really love to do is to use the dictation feature and just yap for like
minutes.
Let's do it.
And then see and tell Steve to come up with a rough plan.
Based on what we aligned on, let's hear your thoughts.
Okay.
All right Steve, we're trying to build a pop-up platform startup.
So we're going to need to think about, well the first thing I want to think about
is what bots will need I suppose.
And there you can maybe work with Dr. Eggbot to help me come up with a small engineering
team to put together a prototype I suppose.
Something that is really loose and scrappy just so we can validate the idea.
So I'm wondering if we just, we don't even need a database maybe to start.
We could literally use like a Google Sheet as a database for that.
Let me just edit this a little bit.
And while Lauren's working on that, I'm going to start my prospecting bot.
So I also installed Dr. Eggbot, really great tool from Lauren.
And I'm going to describe, I need a prospecting bot.
This bot is going to help us evaluate the feasibility of doing research for our restaurant
pop-up business.
So it's going to help me find restaurants and it's going to help me find people to go
to potentially these restaurants.
But we're going to start with the restaurants and the restaurant tours because that's really
big, that's really important.
So what I'm thinking is like maybe there's some way that we can source this from X.
Maybe some people watching the stream like know someone that owns a restaurant in San
Francisco hint, right?
That might be one source but we could also start to think through if we need a landing
page like a website where people can submit their resources or maybe we just need
a phone number that people can call or email that people can reach out to.
So we need a prospecting bot that can help me evaluate these options and eventually help
me build out these systems.
So I transcribed that that's now going through to my, it's going through to my
Grockbot here.
Dr. Eggbot's going to create this prospecting bot.
Looks like we also have chat on the screen.
We have all the bots that we're creating so you can kind of follow along with those
things.
And as we build these things out, as all of these systems get better, we're
going to keep you updated on that.
And is Steve like a Grockbot template or is it just something that you guys
have to do to help you build Dr. Eggbot?
Oh, Steve, no, Steve was just literally the default bot.
I just re-meaned.
But so actually the cool thing about Dr. Eggbot is that, like I mentioned, the
doctor can look through all your bots and modify them.
So for example here I'm actually going to ask Dr. Eggbot to turn Steve into my
chief of staff slash executive assistant because from now I'm trying to think
about, you know, as I start to build this prototype, I'm starting to think about
how to split up the work across all my bots so I don't just have a single bot.
But this lets me, you know, each bot can own its own context and become a
little bit of a specialist and then my sort of team can work together.
So, yeah, I'm going to ask Dr. Eggbot to turn Steve into my chief
of staff executive assistant.
I think I'll just mostly talk to Steve.
And then in the same time I'm going to reply to this message and say,
yes, make a prototype bot.
Make some loose HTML, CSS, GIS, vanilla prototypes, and give it a fun name.
I'm going to make a repo for this.
I'm going to make a repo for this.
I'm going to make a repo for this.
We'll keep this up private for now.
We'll accept PRs after the live stream.
So we don't have to worry about accidentally putting stuff like that.
So here is a repo for this.
We can always rename it if we want.
Yeah, I got the real Dr. Eggbot.
Yeah, I feel like Dr. Eggbot is a must have in my opinion.
I almost feel like it should be a big issue.
I'm biased. I mean, I made Dr. Eggbot.
There's some people in the chat asking if you can download Dr. Eggbot yourself.
And yes, you can.
There's a marketplace for all these bots.
You can go into Grotbot and it's actually right there up front
in the center when you click to add your own bots.
Tell Dr. Eggbot that I don't like that name.
One of the cool things, I didn't realize this at first,
you can change everything in Grotbot just by talking to Grotbot.
So if you're like, hey, I don't like this name, change the name.
Hey, Dr. Eggbot, talk to Steve on Lauren's screen.
It'll talk to Steve.
You can see now it's asking for a new name for the prototype.
And everything is malleable.
Everything is chat first.
That might be unintuitive if you use other tools.
So as you're thinking about customizing the product,
think about these things.
As Lauren's starting to go through this stuff,
I'm going to start thinking through different channels.
Like, hey, how do we want to collect this stuff?
We need some way for people to reach out to us.
If they know a chef, if they have a space,
we need some way to start investigating spaces.
We need some way to start sending emails to these people.
Do we even have email? Can we reach out to people?
Do we need to figure that out?
I think there's a plug-in. I think there's a way to do some email.
Do we have email accounts for the company?
We got to figure out our company email accounts.
We got to figure out maybe a name we can start to align on
what we're going to call this.
And then once we have a name, we get a domain.
Once we have a domain, we get, I don't know,
we get a name for something.
It doesn't have to be crazy.
We need a company name.
We need a company name.
What are you going to call it?
We need a company sign.
Someone in chat suggested potato lab.
Potato lab?
It's about food, right?
Sure.
Potato lab.
Maybe we can riff off on that.
What we're doing is a restaurant pop-up in San Francisco.
Pop-up as a service.
We're going to start with Grog Bot
and start to look through names that are available as well.
Okay, chat. I'm going to need your help
coming up with a fun name for my prototype robot.
Someone said Grog Bot.
We have some options that Steve has given us.
Pickle, natural, waffle, dumpling, churro, meatball.
Can you all spam the chat with your favorite one
and I'll try to squint and see if we can see something?
Oh, this is a name for the bot?
Yeah, this is a name for my prototype robot.
Okay. I'm seeing some votes for potato, churro, nacho.
A bunch of potatoes in there, nacho, Grog Bot,
mashed potatoes, Grog Bot.
I kind of like Grog Bot. What do you think about that?
Yeah, Grog Bot's good.
Grog Bot's good.
And all the pros and cons, like, do you know what kind of restaurant tiers
would be interesting doing pop-ups?
I'm pretty open.
We should definitely ask Grog Bot these things,
but, like, you know, I'm actually chatting with my Dr. Egbot
to build a prioritization bot.
This is overwhelming. I'm on the stream talking to people,
a bunch of people walking.
It's like, okay, how do we start to break down these problems?
Actually, what I use AI for a lot is, like,
I'm just gonna say what's on my mind.
My mind is, like, very active. It's always, like, racing.
And I'm gonna have AI help me, like, slow down,
think through what we need to do.
And Grog Bot's telling me, hey, we need to, like,
find the ideal restaurant tour brief.
We need to, like, do a spike on
what we're gonna start shipping,
what we're gonna start working on.
We're going to, like, start to map out, like, our leads.
We're gonna work through our outreach.
We're gonna, like, get our guests after the restaurant path
has, you know, is formed.
So really, it's like, okay, like, let's break this down,
and that's what's most helpful for me
as I'm creating these bots.
So I got Grog Bot running.
I don't know why.
That's probably a bug that we need to fix,
and I'm gonna, actually, I'll just do it myself.
But so this is kind of, like, how I like to prompt.
I like to get some
high-quality context into the chat first.
So before I even actually start with prototyping,
I'm gonna say, what are some high-quality sources
of good landing page design
and find some references,
and then I can sort of pick and almost maybe assemble,
like, a little mood board
before I start building my own landing page?
A landing page?
I'll be honest, it's been a very long time
since I've built a landing page.
I don't know how scrappy we want to be.
It could be literally, like, Times New Roman, you know,
super, super ugly.
I have a design bot ready to go.
And so if we wanted to do some design polish.
Is the landing page for, like, a particular restaurant?
Or is it for, like, restaurants to sign up?
Well, we're gonna build a platform.
But I feel, and I'm curious what you all think,
but, you know, I was talking about dog-fooding earlier.
What if we ran our own pop-up?
Yeah, we, yeah.
We could do that.
We could build a platform and then also use it to run our own.
Is that crazy?
I think it's crazy at all.
It's too ambitious.
I mean, that feels real, right?
We could actually go find a real place to run a pop-up.
You can just sell some French fries down.
Pop-up platform, and then we're gonna run a pop-up.
I think that sounds like what we're trying to do.
That's pretty ambitious. We like to do ambitious things.
It's like, we might pivot.
That's a good start.
Awesome. Yeah, I like it.
All right, I'm getting a Grock Pot to come up with.
Well, it's found a few galleries already for good landing pages.
So I'm gonna get it to just pick a few directions,
and then I'll see which one looks the best,
and then we'll go with that.
Well, you might want to tell them,
tell them you want to sell French fries or whatever you want to sell.
We're gonna sell French fries?
Let me talk to Steve.
Let me chat again.
Can you tell Grock Pot that,
well, here comes the stream of consciousness.
So bear with me, Steve.
So we're gonna build this pop-up platform
where you can sort of host or
use our software to spin up a pop-up, I guess,
in your city or wherever you are.
And we also want to dog food our own software,
meaning that we'll maybe figure out
how to actually run our own pop-up.
So wait, do we actually need two landing pages?
One for our own thing,
and then one for the platform.
One for potential restaurantiers,
one for restaurantiers.
People buying French fries.
So you think we're gonna connect people?
We haven't figured out what food we're selling
or what the pop-up will look like yet,
but our goal is, okay, well, we need to find people
that can both sell food or create food
or professionals.
At some point, we're gonna need to find people
who want to come to our thing,
but the start right now is gonna be finding location,
finding all these things.
So we probably want to make it look kind of reputable.
Like, okay, here's what we're doing,
and we're gonna have to do some content that's happening,
and then there's also a landing page
probably for this pop-up.
I feel like we're gonna need a lot of landing pages
because we're gonna, don't we need like,
the human nature will need to find like,
restaurantiers, we need to find,
so businesses who are,
businesses or individuals who are interested
in hosting a pop-up,
but also people who we can hire, I guess.
And we can do outreach, right?
I don't think we need to build all of this stuff
to start, right?
So we're gonna start collecting information.
We could like do, before we build software,
it's kind of a manual process.
So we start looking for...
I'm gonna send out this prompt.
Hopefully you got all of that, Steve.
Oh, you were recording the whole time?
I was recording the whole time.
I mean, you know, it's hard to type and talk.
I feel like, Lauren, you can build a pop-up
French fry thing, and then Matt,
you can build a supply site.
Right now, I'm working on some documents,
so I'm clarifying like what we're doing.
And I think this is important for our bots,
because when we have these bots, we're giving them context.
And context is just the information.
And so what we want to do is take the information
that everyone sitting here has in their head.
Oh, my goodness, that's a prompt.
So here's a little trick that I like to do
as well as prompting is like, you know,
you can dump all of this.
Like, so one of my favorite ways to prompt
is to, you know, like just turn on the microphone
and just, yeah, for a few minutes.
And then at the end, I often just say this,
like, resate to me what I just said
in your own words, so I know you understood me.
And that's like much quicker than me actually
looking through what this whole blob of text was.
Let's see what we got back.
So here's what Steve heard.
You're building a platform that helps people
spin up a food pop-up in their city.
Software plus ops to connect the pieces.
You're also going to dog food it
by running your own pop-up with the same stack.
That means more than one public face,
not one mega page.
So we're going to need a platform page
for professionals, rest rotiers,
operators who want to run pop-ups.
Our own pop-up event.
I don't know what we're going to do yet,
but maybe we'll just have a mystery pop-up.
Yeah, well, yeah, maybe our prospecting
will lead us some fun results.
Maybe we find some partners that want to help us out with this.
Oh, but here, Grock was actually making,
or Steve was actually making a really good suggestion.
Yeah.
So right now the bottleneck isn't setting up
a bunch of landing pages.
It's really the first loop.
We're getting our location, getting operators,
and the people who make or sell the food.
And then the attendees come after.
This is true.
So, yeah, this is actually really good.
That was a really good take.
Oh, yeah, we've got some prototype authority.
Let's see what we have.
So basically what you're saying is like,
we just need to do it.
We need to start working on this pop-up.
So we're going to work on that loop, the pop-up,
and then as we work on the loop,
the software that we build for that loop
is basically our company.
So we need to find a place to have this and a chef.
So how do we do those three things?
We'll probably need pieces of software for that.
But why are you assuming that you need tables?
Because when you just do a pop-up thing,
people can come to the counter and buy stuff.
I mean, we could do that.
But I think what I had in mind,
or the way I took over at this,
and we're still working on the idea,
so correct me if I'm wrong,
it would be really fun to do an experience for people,
which is we actually find a chef
or someone that we can have an event.
So let's find a warehouse,
let's find a chef, you buy a table.
There's an experience, and people can come to this experience
and have something.
And that requires a lot of people, a lot of prep.
What you're describing actually is lower lift,
which is like, okay, well maybe we find someone
that's really good at making French fries.
We just do a counter or something.
And then that actually eliminates
a lot of the need for bookings.
So I had like, oh, I want to book this time.
I want to be here.
I want to call in and reserve a table
versus something like a line
where it might be easier to go about that thing.
But I'm not, whatever you guys think is the best,
we can explore ideas.
Yeah, I think it'd be great to maybe take a look
at what we got as some early prototypes
and maybe push this out and see if we can get it deployed.
It's always a cool moment to have something
just live on the internet.
Maybe we can get a preview link somewhere.
Yeah, we have three directions so far.
We have this dark themed one.
And then we have this, I guess these two
look very corporate, I guess.
Yeah, yeah, it's too corporate.
And then the third one, I'm not a big fan of, personally.
So for these prototypes, did Grockbot
make this or did you use cursor?
Grockbot beat it.
Okay, so that's interesting.
What's cool about this is Grockbot can actually just
give you HTML in Grockbot.
And so in this case, we didn't have to deploy
anything anywhere.
This is just like HTML and CSS that just
is being rendered in Grockbot, which is pretty cool.
So we're going to do some feedback, right?
Maybe we can try and adjust some of these designs.
It looks like one of your bots here
that's going to be there.
Yeah, I'm going to do some stream of consciousness again.
While you're doing that, I'm curious,
we haven't got a chance to pick your brain a lot
about this particular type of business.
And I'm curious, when you think about
solopreneurs or people who are creating businesses,
are there any things that you see there
or things that you found to be particularly good advice,
maybe advice that we should bring forward
over the next few days here?
So what I kind of realized as a solopreneur is
you want to design the business
to do more of what you actually enjoy doing.
You know?
Because solopreneurs have a lot of time.
There are certain choices you can make that can make more money,
but then you end up doing things
that you don't actually enjoy doing,
which kind of defuse the purpose of being an entrepreneur.
Yeah.
And I guess, I mean,
a lot of people think a lot about idea generation.
We spent a good bit of our morning today
just thinking about ideas and how to find the right idea.
I mean, you've talked with a lot of folks.
What are your takes?
Do you have any suggestions for folks
on how to find the right idea?
Is it sort of the traditional stuff we've talked about,
like testing stuff and seeing what works?
How do you know what's a good idea versus a bad idea?
I do think it's about getting a market signal
as soon as possible.
Because when we were PMs before the Seoul AI thing,
it was all about having all these internal debates
and making all these crazy documents
and then shipping something
and then you're like, oh, nobody actually wants this.
It's about trying to talk to customers
and get markets signal.
Trying to see if customers will actually pay.
Sometimes they might tell you, oh, this is a good idea.
Just be nice to you.
But try to see if they actually pay.
And don't be too tired to your ideas.
Ideas are cheap.
But the team is what's important.
I guess, I mean, one thing I've noticed
is that it's so much easier to test ideas now, too.
And so what do you recommend?
Is it sort of like try a bunch of things
and see what works or like, I mean,
what is important to...
You can see a landing page.
You want to make sure the landing page
gets in front of the right people.
The people in the chat might not be the restaurantiers
that we want.
But I also think it's also...
Just kind of aside,
I think it's actually like,
anyone can build anything now,
but I think it's actually harder to make money
from pure software.
Because if you don't have a physical component
or a service component,
you're going to be sad to have some sort of landing page now.
Why would I pay for this?
Why would I pay for a landing page
with a database behind it?
So really testing the market,
getting to that first dollar
with maybe something we should try out.
People will pay for the hard stuff.
Like what Gropp suggested,
which is actually finding a restaurant to cook meals and stuff.
I'll pay for that.
I think we're going to try
and lock in for a few minutes
before our next guest hops in.
I guess Peter, before we let you go here,
thank you so much for being here
and just hanging out with us.
Any parting thoughts, any sort of words of advice
will probably be DMing you on the side afterwards.
So we'll definitely keep doing that,
but just kind of like seeing where we are so far
and your view on just other businesses
and product craft like what do you recommend?
I think...
Well, first of all, I joined here
because I really love the product.
I think not on to be too salesy,
but I think this is the first product that I used
and it actually feels like I'm talking to real people
or like I do some sort of people with personality.
It's not just a bunch of chat threads.
So I really like it.
I think my tip for everyone who's using Grockbot
is think about how you can make these bots
more proactive, right?
Try to work with it to figure out something
like tracking landing page metrics
or prospecting restaurant tiers
and then ask it to set up
some weekly crown jobs or schedule tasks
to just send you updates, right?
And I like to...
Simrith Lauren, I have to get a format in a way where like,
hey, send me updates as a number list
so that I can like respond one by one
and say like, hey, this is a good idea, this is a bad idea.
And now I've set up all my bots
so that like every Friday they all kind of checking with me
and I can kind of put my feedback
and I think I'll go off and do work.
So it just feels great.
So what I'm hearing is hire a bunch of bot employees,
get them to automate the stuff
that you don't want to spend your time on necessarily,
but then maybe get them to be proactive
so that they're kind of pushing things to you
versus you having to redirect to them.
Exactly, exactly.
And like a quick feature request before I go.
Okay, first feature request.
Yeah, I think it's important.
Like, why do you guys even have to use Slack?
Like if you can let humans jump in this interface
to talk to the bots in channels and stuff.
You know, like the thing that kind of feels lonely to me right now
is like, I'm just talking to my bots.
I want to talk to Lauren. I want to talk to other people.
Right? And like,
I don't want to collaborate with each other.
Maybe if it's coming soon.
Okay, bot collaboration. It's in the feature queue.
We'll get it logged.
Human bot collaboration.
We'll get it logged. I think we'll probably see some need for that
as the three of us cook with our own bot teams today.
With that, Peter, thank you so much for hanging out with us.
This has been awesome.
We will definitely see you soon.
Awesome. I'll be buying some French fries now.
See you guys.
Yeah, thanks.
So for those catching up,
I am making, so I just,
well, we did a couple of things.
I gave Steve some feedback on the prototypes so far.
A little bit too boring.
A little bit too bland.
And it messaged GROCPOT to,
oh wait, I typoed,
it messaged GROCPOT
to come up with some new prototypes.
And then in parallel,
I'm also asking Dr. Egg Bot
to make us a cracked engineer bot.
So GROCPOT is our prototyper.
And then I'm going to have
a virtual engineer bot
to start thinking about like
tech stack, you know, how we're actually going to build things.
I'm going to keep the prototyper
for just quick throw away prototypes.
And of course,
I need help naming the bot again.
And I think we have some good ones.
Chat, I need more help.
Give me some good suggestions.
Chat, we got to lock in.
Spud. Spud is pretty basic.
Spud or Tater.
What do you guys like? Tater or Spud?
You like Yeogi?
Kind of like Tater.
Let's do Tater.
That's the potato override right there.
Potato gets the point I say.
So we have some more fun prototypes.
Oh my god.
Is this too fun?
The designers are rolling your eyes.
This one looks kind of like
a receipt I guess.
I think one thing maybe to call out is like
when GROC,
when we talk about like models,
when we're just doing the work here,
GROC is the model and then there's a harness
which is the code that runs the model.
And GROC Bot is very lightweight.
Like you might have heard of coding
harnesses like GROC build for example.
But GROC Bot is not that.
So these are very lightweight prototypes
and when we do more serious engineering work here,
what we're going to do is connect cursor
and then we're going to kick most
of the engineering work over to cursor.
And that's because the cursor harness
is very good at writing code
and so as we start to build this idea
what you're going to see Lauren do
is start to rely much more heavily on
cursor cloud agents and we'll walk
through some of that stuff as well.
But that's a good way that like as we go
from like low fidelity to high fidelity,
you'll start to see some of these products
evolve and that's completely normal.
Yeah looks like we're making some
engineering decisions too.
Yeah I'm going to say
Oh wrong bottom.
Oops.
Go to Steve.
I want you to ask Tater
to start thinking about
our tech stack for building out
this platform.
I think we have some friends coming
from Griselle
and PlanetScale maybe.
Spoiler alert.
So we want to probably use
those options.
But in terms of other
tech stack I don't know I'm not
too stubborn right now.
So just give us some good options.
Yeah I kind of feel like we should let
the bots just pick and get something
put together pretty quickly.
Obviously avoid endless tech stack
debates and as we
seek product market fit here.
Yeah.
Awesome and while that's working
literally all I've been doing on my screen
is writing out a Notion Doc.
So I know Lauren said she doesn't like Notion Docs.
I find it very helpful for planning things.
So you know on the A.B. side if we cut
to my screen kind of all I'm doing
is defining essentially what we've agreed on thus far
which is like our company name
is like ship by Thursday and we're shipping
by Thursday until we come over
with a better name that more clearly
elaborates on what we're doing.
And so Lauren I think is the CTO
here.
She's really the most technical.
You're running tech.
I think Rochon is the CPO
he's the head of product.
He knows all the product details.
The chief product officer.
Not chief potato officer.
Lauren is the chief potato officer
of the CTO.
I'll highlight that in our company.
Can we get like some like titles?
Yeah we need some like placards.
I have some paper like here.
I can't write and talk at the same time.
Lauren's going to work on some like official titles here
for the company.
And what we're doing is we're building a restaurant pop up in San Francisco
for the experience.
Well actually we haven't decided it's a restaurant not
but we're going to build a pop up in San Francisco for food.
We're going to use that experience to build an OS
for running pop ups this meta product.
And so our goal is to create a business that allows us to do
more of what we enjoy doing which is bringing people together.
And so in this doc
I'm just saying this is coming together
really quickly so we need to buy a smoothing fast
and getting stuff done.
And there's a bunch of stuff we need to figure out.
So I'm going to kick this over to like all the grokbots I use
so they at least have sort of
a context of what we're doing.
And the main thing I want to do
is keep that context consistent.
Because as we bring guests in the show
I have to explain to the guests what we're doing.
I also have to explain to my grokbots what we're doing.
So yeah we're cooking
and I think we might even have another guest
coming on pretty soon.
Are you our CEO?
I guess.
If somebody has to do it
I won't be the one to ask him to do it.
You can be the grown up.
Wait what is the e-sense for executives?
Exactly.
I'm sorry I forgot.
Oh my gosh.
Outstanding. Chief Tater Officer.
Chatt is crazy.

You know, I was talking with like some of the very official name plate.
Thank you.
What do I do with this?
Like what do I?
What's your Twitter handle?
Fold it out.
Atrociousn underscore s.
Roushn.
Lauren, what do I do with this?
I don't know.
Just like stick it somewhere.
We're going to work on that.
We'll work on that off line.
And I'm told the team is working on bringing in a virtual guest.
But yeah, I mean, if you want to cut to my screen,
we can start doing some like planning here.
I like to plan things out.
I like to prototype things.
I love it.
Thank you.
Thanks, Lauren.
Very important.
Titles.
We'll have to get those tinted up and properly made.
I'll do some origami.
Does this work?
Lauren, the chat has noticed that you
have an update available in the Grockbot app.
Oh, man.
We'll send updates for when we're talking.
I don't know if I'm brave enough to do that on stream.
Yeah.
We'll do that after.
We should show you.
We'll do that in between the break.
How fast the team ships.
It's kind of crazy.
Oh.
We have a few more colors.
I don't know.
Maybe we don't have to make it look that good.
OK, CEO, please.
Make an executive decision right now.
Which of these three?
Prototypes don't matter too much.
Let's go with pink.
I'm feeling a little crazy.
Night market pink.
Let's just do that.
We're going to roll with that.
And then these don't matter so much in my mind.
So let's move fast on this.
And eventually, we'll get cursor cloud agents hooked up
and we'll find some design help somewhere out there.
Yeah.
Hey, design bot.
We got to get a designer bot.
I feel like that's going to be really great.
Just a level set for the folks watching and maybe
for our team here what's going to happen in the next few minutes.
We're going to have a guest joining us in two minutes.
I think what we'll try and do is maybe get some builds
going in the meantime.
So what do you guys think we should kick off?
I think it might be great to kick off some cloud agents
to get some front end and back end stubbed out.
Should we get some code in a repo?
I guess, Lauren, what do you think we should maybe
kick off?
And we can build while we're talking chatting with our guest
and make some progress here.
Yeah, do you want to set up the GitHub repo?
Yeah, I just made one, actually, and put it in Slack.
But it's just popup.git.
Can share that if you have it.
So this is our Slack.
Yeah, it's our Slack channel.
Well, look at my name.
I should probably set a real name.
Lauren Chief Potato Officer.
I'll put a picture later.
Where is it?
It should be right there.
OK, OK, so we got our, I should probably zoom in too.
Go, go, go.
OK, so let's try and get the landing page built
and pushed through this repo.
And then start figuring out how we can host it.
Do we want to buy a domain name right now?
I think I have a Vercel account set up.
So maybe we try with a test account,
then we can always grab a domain once we know we want to call this thing.
All right, sounds good.
You know what, I'm going to ask Steve to give us some domain names.
Yeah, that's a good idea too.
Or we could get, we have a marketing bot.
Maybe we hire a marketing bot.
We should hire a bot to think about how we should brand things,
name things, then coordinate with our team.
Since you're the chief product officer,
do you want to figure out the domain name?
I'll generate a few, but I don't know if I'm very good at naming things,
as you can probably imagine.
Well, most of the more things we just start to get stuff up.
So like, yeah, let's start ripping.
Yeah.
Cool.
In the meantime, I'm getting bot cooking right now.
So we can, and what's going to be great,
I think once we get some code into a repo,
is that all of our bots can start making PRs against it,
as we can kind of all evolve together.
This is important to me.
Thank you.
We got it.
Oh, no, my name plate fell over.
These are the important parts of running a company.
We'll get some more official ones.
The team is working on the infrastructure here.
Yeah, this is like company building exercise.
OK, we got some ideas.
I don't know if I like them.
Oh, what?
Oh, that's actually pretty fun.
We have claimstall.com, stall.run, nightmarket.app, popup.place,
openstall.co, peanutpopup.
That's kind of cute.
Peanutpopup.com, pappeto.com, and hostapopup.com.
I don't know if I like any of these.
Do we have a Notion doc that's sort of like a version state?
Do we have a Versaille account already?
Can you add me to it?
Yes, I can.
I'm going to send you the company doc version.
Oh, great.
Yeah, I want to feed that to my bots.
They have context on what we're building.
It's very bare bones.
I like to bias, even if we cut over to my screen.
I think simpler is better.
And if we think about all of the stuff that goes into AI,
the more information you give it,
the more it's going to lean on that information.
So if you give it a bunch of bad information,
you're going to need a bunch of bad outputs.
And so what we want to think through is like, hey,
how do we give the simple, most concise description of what
we're working on to AI?
And that's what I'm trying to do here.
Send that doc over to Rotion.
And we're going to keep moving through it.
Cool, cool, cool.
What's the best email for you, Lauren?
I think you should have one.
It's the one in my Slack profile.
OK, let me grab that.
Members, I'll make you an admin on this.
Do we have cursor set up in here?
We should add the cursor bot.
Oh, yeah, we definitely should.
We're going to need some.
Nor Slack.
OK, copy, mail address.
OK, Lauren, I'm adding you as an owner to our Vercell
organization.
OK.
And I have not set up Vercell on the repo yet.
So as soon as we have a repo, we'll probably
want to hook some stuff up.
Oh, wait, I thought you were setting up the repo.
We have a repo, but it's an empty one.
Oh, OK, OK.
I think it's already live.
Steve did it while we were talking.
I love it.
OK, well, while we're waiting for our guest to show up,
why don't we do some work here?
I don't know.
Yeah, it looks like we're switching.
Great, we got some screens up.
Extremely.
That's our platform.
It's an index, not HTML, but that is more than we had
when we started.
So let's see here.
Wait, the styles are all in line?
OK, amazing.
I love that.
This is super scrappy, but that's what you need.
I'm going to just go ahead and get
Vercell connected up to the repo here.
Hit the deploy button.
Yeah, I need to authorize my new Vercell.
Configuring GitHub here.
Got my email.
Cool, OK, yeah, I see our pop-up repo.
OK, I need to actually.
All right, we should get some deploys going soon here.
Oh, I need to sign up for an account first.
Sign up.
So yeah, while Lauren's signing up for some accounts,
I'm working on creating a couple bots on my side.
I feel like we just need to start cranking through
some things here.
So I'm working on an op spot and a creative director.
The creative director I'm thinking is going to help me with naming,
high level details, some of these other things.
And then the op spot, I just need to start doing stuff.
So I'm setting up Slack.
I'm getting some other things.
Cookin'.
It's kind of everything you do running a company.
This is like, obviously what we're doing is super exciting stuff.
Humans have to do this stuff.
And we're going to start to move towards automating some of that as well.
Yeah, I think it's really just cranking.
Let's see.
Oh, I thought I added you.
There's a pending invitation for you.
How do I?
You might be in your email.
I'll check.
I can also see if I can invite you by link.
Oh, I see it.
Awesome.
How am I going to do this?
Can I hide my screen for a second?
Yeah, can we cut the screen please on the AV side?
I can stop streaming.
Yeah, you can cut that.
It's fine.
Oh, that was my face.
Oh my goodness.
Don't worry about it.
They got us covered over there.
OK, OK.
We are down on the stream.
Oh my goodness.
No, can somebody else stream?
Yeah, I think I can hop in.
I got it.
I got it.
Little stream rescue here.
Don't worry.
I need to.
I mean, I'm also streaming.
I just don't think that they're picking up the ball.
Oh no.
Don't worry about it.
We're live on the mic, so we'll be right back.
Technical difficulty right now.
OK, so I'm connecting Notion here.
And we have.
All right, we're back.
You can see us.
All right, yeah, we are still cooking.
OK, OK.
I'm off now.
OK, let's see how we're doing on our deploys.
I see a deploy, guys.
Oh man.
All right, hold on.
Oh, did you deploy it already?
I deployed.
Nice.
We have deploys running on main, so just push to main.
Hop up orphan omega.virtual.com.
There we go.
There you go.
Wait, but I guess right now it doesn't do anything.
It's just an HTML page.
We need it to go somewhere.
Yes, we need to go off.
Should we just collect these in a Google Sheet
somewhere?
Where do you want to collect our emails?
I think that sounds good.
Yeah, but how will it be easy to connect Google?
Maybe we should just dump things into a.
Oh yeah.
Like a.
We could just use an Notion database.
Yeah, we have the Notion MCP set up, right?
I don't know if Notion's going to be able to handle
the load from everyone.
Dump into a flat file database.
OK, I just connected my Griselle account as well.
Cool.
Awesome.
OK, then I'm going to ask you.
I'm going to say, OK, Steve, we've deployed the landing page
already.
But obviously we need to connect the landing page to actually
store the signups somewhere.
So for now, could you connect it to Notion
and make a little Notion database for us?
And then I guess you'll need to add me to the Notion.
Yes.
And by for.
In the meantime, I guess we should
start looking at the tech stack choices.
Host on Griselle, yes, makes sense.
App.
Could we?
Actually, maybe crazy idea.
Like what if we just integrated PlanetScale right now?
Should we just go straight to the real database?
Because I guess if we do a quick test,
we're going to have to rip it out later, right?
Yeah.
Yeah, let's do that, actually.
That might be cool.
And once we get it wired up, then
we're kind of off to the races.
I wonder if that would be easier.
I do want Postgres instead.
So let me say a small tweak.
Let's use PlanetScale Postgres instead.
Payments later, that's true.
Must have.
Nice to have.
Yeah, we don't need any of these nice to have stuff.
And we.
OK.
Anyone can use the box?
I'll just give a feedback on this plan.
And then I'm going to go back to Steve and say,
connected to PlanetScale.
I'm going to need to be added to that, too.
Oh, yeah.
I think we have a PlanetScale work set up somewhere, right?
We're working on that.
We can reach out to the PlanetScale team this morning.
So I think we should go somewhere.
Let me get this up.
I'm not stirring the screen right now.
OK, sweet.
Yeah, we'll get the PlanetScale stuff cooking.
Thanks.
Shout out PlanetScale team.
I think we might have some guests from PlanetScale
in the next couple of days.
Love it.
Steve, roll with it.
Sweet, sweet.
And for those, we're figuring out
how to get this screen sharing back up.
But for now, just to give you a picture,
if you're just joining the chat of what's happening,
we have an ops team cooking over here with Matt,
putting together a company playbook,
and making sure we have a go to market motion.
We have an early landing page prototype that is just an HTML.
We're leaning towards a pop-up business or a pop-up
platform where we make it easy to run pop-ups.
And maybe we run our own pop-up, too.
And we have Lauren cooking up some integration work,
I think currently working on getting our database set up.
That's the recap.
You know what?
I'm going to say, yeah, I need the PlanetScale account.
And then I can tell it to set it up for us.
PlanetScale is that we have accounts.
Shout out to the PlanetScale team.
So we'll be good on that.
We can keep moving there.
Cool.
PlanetScale, let me find the invite and get it set up.
I think we still need a domain name, right?
We still need a domain name.
I got my creative director.
Let's.
We can also just as chat.
Chat.
Give us some ideas for our domain name.
We need some fun, cool domain names.
And also, we need to start.
We need to screen share again, I think.
Ooh, I wonder if ShipByThursday.com is open.
Is that domain open?
ShipByThursday.
Yeah, unless someone sniked it from us in the chat.
I should stop.
I should stop.
You should probably, if you want to,
buy the domain right now.
Yeah, we might be able to buy it through Bercel
and just have it tied directly to our project.
Cool.
I'm hearing that our next guest, unfortunately,
won't be able to make it.
So you're just going to get more time with us building.
I mean, we're kind of busy.
I think we've got to cook here, man.
I feel like we're falling behind.
Sweet.
When I'm going to say.
OK.
And hopefully we can get more screen up in a second.
So I'm still reading on the client-scale credentials.
But I guess try to unblock yourself as much as possible
and think about the table design and maybe set up
some dependencies for us.
But just, yeah, unblock us while we wait for a client scale.
Can we turn that off?
No?
I think we're showing the wrong screen right now.
Sweet.
All right, we're good.
We have our cell accounts, right?
Yes.
Well, I don't know if I've invited you yet, but let me do that.
OK, so exit.
We'll get there.
What, you're using the email for your same email you sent?
I'll use my, I can use my personal.
We have no screen on.
I'll slack you my personal.
And then by the way, Steve, I want you to tell Tater
to use Cloud Agents going forward.
Maybe make a project agent.
So, yeah, for those of you watching,
Grockbot is a really powerful tool for orchestrating bots
and obviously automating a lot of things.
And then Cursor's harness is really good for coding.
So one thing that I like to do is to have Grockbot
basically spin up Cloud Agents for me.
And Cloud Agents are agents that run on Cursor's harness.
So it has all of the smarts and the engineering know-how
built into the harness that Grockbot can manage them for you.
So we can actually do all of our engineering work in
Grockbot, even though we're using Cursor Cloud Agents.
And also the nice thing about Cursor Cloud Agents
is that they run in the cloud and they also
have their own virtual machine.
They can test our application.
They can click around.
They can take CPU traces, all that good stuff.
And so we can do that natively right here in Grockbot.
All you need to do is tell your bot to spawn a Cloud Agent.
And it can also, if you have an existing repo,
you can tell it to help you set up your environment
so that your Cloud Agents actually
know how to bring up your application
and run it in the cloud.
So what I've done so far is I've told our engineer bot, Tater.
Maybe I will give it a title.
Can I still give it a title?
Maybe not.
The label, maybe?
I don't think we can add a label.
OK, so Tater is going to use Cloud Agents now.
And you can see that it's created a new project for us.
And a project is a little hard to describe.
I guess it's like a special kind of Cloud Agent
that is really good for long projects.
So I'm telling you to spin one up.
And we'll use that to keep track
of all of the technical engineering context
that we have for our application.
And just whenever it needs to do more engineering work,
it just sends more messages to the project.
But we'll see how that goes.
Maybe we won't do a project.
We'll see.
Do we have the planet scale credentials?
Yeah, I'm working on that.
Let's see here.
I mean, we can run it locally too.
I'm going to say.
We're still waiting on the planet scale credentials.
So maybe let's for now do all of our development,
I guess, somewhat locally on either your cloud machine
or in the Cloud Agents machine and come back to me
with a video or a screenshot.
And we'll worry about deploying it later.
And on my end, we now own the domain shipbythers.day.
I like it.
So we're going to get some stuff going on there.
Do you want to wire that up to our deploy
so people can start looking at the?
It's on Vercel.
So I was going to get a company lander.
Like, we need the ship by Thursday company lander.
And then the product, I think, can go to an actual product domain.
So ship by Thursday is going to be the main site that we're
ship.
Or it's our company.
It's the meta.
It's the company behind the companies.
And then we're going to start working on actual company.
We need a product name.
What is this going to be called?
We can work on that.
We've got plenty of time.
We need a working title.
OK.
Like, I don't know.
I didn't know the thought.
So now I'm also, because we have a domain,
I need to deploy some code to it for our company lander.
And what I'm working on is using Dr. Eggbot
to install our cursor team kit, a bunch of other skills,
P-Stack, and then connecting that to our cursor cloud agents.
When we have that ready to go, I'll
be able to start writing code to get that up there for y'all.
And maybe making things look nice.
But yeah, we're just hacking, right?
So you're here with us as we're building the company.
We really got a lot of stuff going on.
We have so much things we have to build.
But that's kind of part of it.
That's the fun.
All right.
I see Steve has made a mistake because he's
opened a pull request.
And I'm going to say, new rule, no pull requests.
We ship to Maine for now until somebody yells at me.
All right.
There we go.
Let's see what it opened anyway for the pull request.
It's got, wow, 2,000 lines.
I'm not going to read all that.
So one thing I would like to do as well for my setup
is to start thinking about the engineering workflows.
So we have Tater, who is our engineering bot.
But I also want a bot that is like a reviewer, I suppose.
So I'm going to tell Steve.
And we're actually going to cut to a hold screen briefly.
And we're going to come right back.
And Lauren's going to pick up with the factory creation.
So we'll cut.
And we'll be right back on some technical difficulties.
Mike, Mike, Mike.
I'm going to make a new channel called.
Hi, Lovell.
We're just getting started building our company.
So we are kicking off our first agents.
I'm working on a lander for our company.
And I can actually see in cursor some of the other agents
that my teams are working on.
Oh, and we have a guest.
We have a guest.
Cody, hello.
Thanks for joining us.
Well, what's going on?
It's fun watching you guys build live.
Just business buying and selling marketplace online.
And then we help people scale businesses
through our advisory business.
And then if you've seen the app on X or the internet,
I think there are about 15 million people
that follow us online.
So we're on the interwebs all the time
talking about stuff like this.
That's awesome.
And I know you've been using Grockbot a little bit,
or at least your team has.
I'm curious, how's that been?
What have you been using it for?
Maybe share some of your favorite.
You've got to be very useful.
So I thought I'd show you a couple things that we're doing.
I think you could maybe see my screen share live.
Can you see something that looks like a dashboard
with a book on it with 135,000 users on it?
Can you guys see that?
Wow, we see it.
Yes.
OK.
Anyway, so basically what this is,
is we're launching a book this week in two days.
And it's a little bit like a product launch.
You can kind of see we're tracking.
We built this with a mixture, a little bit of Grock.
This is located on Bursell.
And basically tracks real time attendees.
And then if we could scroll down, I
didn't know what our capability was going to be on here.
You could see what's paid versus organic, what channels
are tracking above, what's our average
cac to average order value for each of the people who come in.
And so since we talk a lot on the internet
and build these products, I thought
we could show you some of the things that help us build.
So that Bursell is one of them.
Maybe the other thing that would be useful,
I don't know if you guys can see something now that
looks like a dashboard.
This is Pulse.
So basically.
You see that there, yeah.
OK, cool.
So this is how we build businesses of those companies
and pitch.
And so they don't know how once they get somebody's eyeballs
on it to pitch well.
And so then you put together a plan based
on your lowest profit lovers.
So anyway, I'm not sure if that is useful or how to.
Let's stop sharing.
But those are a couple of things we actually built with AI
and how to figure out what business in particular you want
to start or what's your biggest problem.
I mean, good question.
We'll always give you the best answer.
Fair enough.
Well, I think we have an idea that we think might make sense.
How do we get some validation for it?
Yeah.
And so if you haven't been following along like our idea
is like we're going to run a restaurant pop up.
So we're going to need to connect the restaurant
for with guests who want to experience that.
We're going to find a location and do some of those things.
And then through that process, we're
going to turn that into a piece of software
that we'd like to sell to people as well.
OK.
And so the biggest issue now, theoretically,
you're going to be able to do licensing, pop application,
leasing, all of those logistics.
We're like, we've got that figured out.
But we got to make sure that we get butts and seats.
And people actually show up.
Is that right?
I mean, yeah, we're going to focus on the ops first.
And then we're going to reach out to legal and figure
out how to make all that other stuff work.
OK.
And so you do want users, or you
want to focus on getting the leasing.
Basically, what I would say to you is my belief is always,
before you even decide what business you want to build,
you should see if you can sell it to three people.
Unless, especially for a restaurant pop up,
you guys aren't going to get venture capital funding for that.
Only 2 out of 2,000 companies gets VC.
So that's all good.
It's a good business.
But what I would probably do is I
would be obsessing on how to get distribution.
Obviously, you have 430,000 people on here.
You have a super unfair advantage.
But let's pretend that wasn't the case.
What would I do?
One, I think X is an incredible place to do it.
I guess I have a couple of thoughts on there, which is one.
And the man himself doesn't seem to be bothered by this.
But I think a little controversy is OK.
I actually pulled up some of my most, let's call it
my most controversial tweets so you guys can see those
for when I get yelled out of the internet.
So basically, I would play around with your controversial tweets.
So this one here, I don't know if anybody saw it online.
It's like do everything in your power to move to Austin.
It's basically an AI generated video.
This was made with Higgs Field.
I think it probably took an hour and a half.

half of one of our editors' time, minimal credits,
and we got roasted on the internet about this
because there's a scene where there's somebody
with a surfboard in it in Austin.
But if you look at the view count, who cares?
Because you can see 2.4 million views.
I was trying to get people to come and apply,
so we have a portfolio of a bunch of businesses
and I wanted them to apply.
We received hundreds of people applying
and then you can see in the comments
hundreds more sharing it.
And so I think the first thing, if I was you,
I'd be playing around with what's controversy.
What do you guys believe that other people don't?
And yeah, I think you guys should make an X account, for sure.
I also think one of the other things you should think about
is what is going to make people talk?
We did this other one where people were really mad
because we said, don't do paid ads to start your business,
which I stand by.
But we made it slightly hyperbolic
and don't do ads until you hit five or 10 million.
So people lost their minds.
You're an idiot, Cody.
How could you think that?
Every business needs ads, right?
But it got a bunch of people talking
and what I really wanted to do
was get people to the book launch.
And so that one got like 5.7 million views
and I believe signed up 2,000 users.
So it's cheap.
So I would start that for you guys.
So what's your little bit of controversy you can do?
That's really good advice.
Actually, and it's funny you mentioned the ads
because we definitely were thinking about running ads
and I guess that would have,
maybe we should reconsider that advice.
Well, it depends.
Like do you wanna,
I think the problem with ads right up front
are you don't really know your target persona.
Like you don't exactly know who you wanna go after.
You have to spend money then to make it.
If you're not getting funding already,
why would you do that?
And three, you can put your ad spend behind something
once you know the organic works.
So like why put ad spend behind bad ads?
Cause you don't know what good looks like yet
until you've done them on X.
That's at least basically how we've grown
almost all of our businesses to start.
Then once you know what good organic looks like
and who your perfect persona is,
you could throw a bunch of dollars on top of it,
which is what we do.
That makes a ton of sense.
Yeah, we should, I feel like we should be
recording this or feeding it into a Grock Bot
so that we can use it to run the next phase
of our business.
Cody, you talk.
Well you could also like take the top performing tweets
from everybody like me who markets a lot online
and feed it into your Grock Bot
and then just feed your Grock Bot.
Hey, we're trying to achieve X result.
You can connect tweet 100 to X,
have it feed all the top Twitter posts
and then say pull me the top 15,
like steal the frame, not the exact verbiage.
Right.
And then retweet them.
Yeah, I really like that.
I wanted to go back to something you asked earlier.
You were kind of sharing some contrarian takes
and kind of leaning into that.
I'm curious like what's the worst business advice
that you've heard given to businesses?
Cause I think, you know, obviously we're gonna have
a lot of folks on, they're all gonna hopefully
give us A plus advice, but it's also good to know
what to avoid.
And so I'm curious just like, yeah,
what's the worst advice you've heard
and what are the things that we should be aware of there?
Yeah, I mean, the advice that I heard in the beginning
and followed was hire A players
and get out of their way.
It sounds really nice.
I think you could say the same thing with bots.
Like, yeah, just let your bots run,
get out of the way.
You guys know firsthand,
cause we run a bunch of them here too.
You can't really do that.
You have to be constantly iterating and overseeing.
And so I don't know how many people on here
run single person businesses
as opposed to multi person businesses,
but you know, we have 150 employees in this company.
And the only time we almost ran out of money completely
was I followed that advice.
And so I think if you're gonna hire great people,
you have to oversee them intensely.
We say hire great people.
Be really clear on what you want.
Incentive align them to make sure every dollar
that they make is tied to a dollar that you made,
and then have an operational oversight.
So with Brock, you could probably build
a little bit of a prototype like we do
where you can track what is my cash on cash return?
What is my AOV?
What is my LTV?
And if you're gonna hire a marketer,
track all of that to their actual incentive aligned goals,
which is what I would do.
That's awesome.
That's awesome.
I feel like there's so much that gets shared conventionally.
And I'm curious for your take on this as well.
You mentioned earlier,
we have a bunch of folks watching the stream.
Thank you to everyone who's tuning in,
obviously like really heartening to see the audience here.
And you've built some pretty incredibly large businesses
and marketplaces.
And I'm just curious for your thoughts
on sort of distribution and if distribution is a mode.
I think you've talked a little bit about this
in your posts as well recently on X.
Yeah, I would love to know like kind of your general
thoughts there, any advice for us as we build,
but also just some general takes.
I mean, first of all overall, like, yeah,
I mean, people hate that take.
They hate the idea that distribution is a mode.
And then, you know, I was talking to Jack Dorsey
and in the XDMs.
And I think he's one of the OGs that says
it actually really, really is,
especially in this world where products getting easier
and easier to build.
I think, you know, if we invested
at Contra and Thing Capital,
I think we've done 36 investments today.
We see a lot of AI investments.
And the number one reason they don't make it
is because they don't have distribution.
So it's not popular.
People won't like it,
except that's what matters to get funding.
So, and then, you know, for most businesses,
you need cashflow and you need distribution
because you won't get funding.
So you're going to need your cashflow to fund it.
So I would say like obsess on that.
And I don't think enough startup founders do that.
And so in the beginning, it's all about building product
or even these days, people say they're building it in stealth.
You guys are doing the opposite
because you're building in public, which we do too.
But then I think maybe the only other thing
I'll show you guys is like,
I would, nobody believes anything, right?
So you have to have proof.
So we like to share on X, you know,
there are always going to be people that don't like you.
There will be people that don't think
that you're doing a good job.
And there will be people that just like,
don't want to see you win.
And so I hope you guys are taking screenshots
of all the stuff you're building with Brock right now
because you'll want it in like six months when this works.
And you'll want to show like your user growth from day one.
You're going to actually,
we have something called a proof vault.
I highly recommend it, which is like every single thing
somebody said about your product that's good ever.
And then we break it down from like testimonials
to video views to screenshots to recently added proof
to has it been verified or not?
Is this like publicly allowed to be shared?
Because eventually people will want to know all of that.
Makes a lot of sense.
If you're open to it, actually,
we'd love to do a bit of an interactive exercise
and maybe help kind of work with you
to maybe create a bot together and kind of prompt it initially
and kind of take some of the advice
that you've shared with us
and turn it into something we can use for our business
as a sound like it'd be fun.
Let's do it.
All right, let's see.
Let me get maybe Matt's or my screen up on this stream.
If not, then maybe we just work in the background here
and we'll share the results afterwards.
Okay, so this is my screen as your screen.
Maybe you can walk us through what we're doing here.
Let's do it.
Cody, we're going to create a bot.
And honestly, I think we'd really benefit
from sort of the marketing and distribution side of things
because we're kind of flying a little bit blind there.
And so yeah, we'd love your suggestions
on how we might start prompting
and Matt, I guess if we create a new bot,
maybe we can give it a name.
Yeah, start.
Yeah, we're doing a new bot.
Well, let's start first with what do you want it to do?
Before we build anything, we got to ask ourselves,
what's the number one thing that needs to be done
in the business in order to make money or get users?
Or what's our number one goal?
Is it getting leads?
Is it getting people to pay and sign up?
What's the goal that we need to hit?
Yeah, so to start,
we're going to try this pop up experience.
So I think you made a good point,
which is like, hey, we want to get three people,
maybe that would be willing to pay us
to prove out that this is a good idea.
So it could be leads, it could just be a sign up form
where we start collecting emails of people
that are interested and then we try to collect
our first signups from that group.
It could, I think getting the word out
or trying to broadcast what we're doing
is maybe a good start.
Yeah, I like it.
I always like captured audience.
So that's probably where I'd start.
I think you're nailing it with emails
because then you can always come back to them.
You probably are not gonna have your,
like everything isn't set up right now.
Is that you don't have the ability to click to purchase,
you know, connected to a reservation system
to click to reserve a spot, none of that yet, right?
Exactly, this is correct.
Okay, so I would start with lead magnets in particular.
You already are good at X.
Almost everybody on here is probably playing on X.
So I would start with a really good lead magnet.
I think usually when we go to start a business,
we go imitate, iterate, innovate.
So who already has this type of business
that you want that's done it well?
Do you have any market knowledge?
Yeah, I have no market knowledge of pop-ups
as a matter of fact,
and that's a good point to bring up.
Okay, well I would run a separate bot on that.
Go find what city you're gonna do it in.
San Francisco.
Okay, well you're gonna have a lot of empty restaurants
in there, so that'll work out fine.
You probably won't have expensive leads.
So you're saying basically what we should be doing
is like background research on other businesses
that have done the same thing.
Exactly, like if you haven't done this before,
go try to figure out who the best person is
that's already doing it and reverse engineer
what they're doing.
Okay.
So that would, the very first bot I would make
would probably be that,
like pull me all the top restaurant tours
in San Francisco,
and then grab me five of the best websites,
then give me the five best pop-up pages
on their websites,
then give me the five best pop-up restaurants done
in any major city in the last six or 12 months,
then put those all into a document
so that you can see them.
And then you're probably gonna wanna figure out
how should we actually,
like what is your branding
and what's your differentiation here?
Do we know anything about
how you're gonna position this pop-up?
Awesome, yeah, no, and I'm tracking here.
So I'm using, we have basically a bot
that can create other bots.
This is something Lauren has educated us on.
And I'm saying, hey, we need an operator research bot
and kind of took some cliff notes
on what you said there.
I'm gonna say, hey, go find the top
San Francisco restaurant tours,
the best sites for those restaurant tours,
similar pop-ups and similar companies,
and then use our sort of like founding document
in Notion as grounding.
Oh, I like that.
Because I've been taking some notes here
on like just the general sort of like,
hey, what are we trying to do here?
What is the goal?
So that bot will be grounded in our business.
And so hopefully, Cody, thank you for the tip there.
This is a good way for us to understand
what we're getting into.
Because usually, right, it's like,
sometimes when I approach problems like this,
even in writing code, it's like,
somebody's probably done this before, right?
And so if we can start from what everybody else has tried
and maybe failed at, we can give ourselves
a bit of a head start.
Yeah, plus your storytelling in the beginning.
So like really what you wanna do,
you're gonna have to sell the dream
in anything you do before you even get the lease
or anything else, especially if you wanna get it for free.
So if you wanna get free things to test your idea
before having to pay for it,
then you've gotta sell why somebody should help you out.
So I think what you're trying to do here is,
and I might just cause I have some context,
if you do, anybody who's building
hopefully isn't picking a random thing,
throwing something at it with a couple ideas in building.
Typically, you're gonna wanna have lived the problem
for a minute, but you might wanna put in here,
how would the best restaurateurs like Danny Meyer
position a pop-up just like he did with Shae Shack
in the very beginning?
How can we replicate that?
And then you might say,
what would a killer landing page
with a one-line title hook
and then some sort of exclusivity, scarcity,
timeliness and relevance,
like the four reasons people do anything,
what would those be for this?
And I would feed that all into something
that'll give you a bunch of outputs.
And this is perfect for quote unquote controversy.
Like what would make people talk about this
in San Francisco?
It's probably something,
I mean, maybe it's an AI restaurant pop-up cafe.
Maybe you riff on why claw bots,
like why claw bot meetups went so viral in San Francisco
with the little lobster hats.
How do we replicate that?
What would be a different version of it?
We have some cute rock bot logos maybe
that we bring into it.
Yeah, yeah, yeah.
You could, my only thing on logos though
is that's about you, not about them.
So everything you sell,
like that's good for your boss and Grock,
but for the user,
it's like what is actually gonna make them want
to participate in that?
That's why you'll see on our ads.
So I think we spent a couple million dollars in ads
for this book launch up,
we'll tell this week, which is going down in two days.
And the ads where it's overly corporate,
like our logo is all over it,
those perform at about 40% the rate of non-corporate ads.
Like our best performing ads
are me and my husband talking to each other
or us inside of our business
or literally one with our dog in it.
So yeah.
So if you're building ads.
So if you're building ads.
Is that fair?
Yeah, less corporate the better,
less produced these days the better.
We did a couple of ads that are really overly produced.
No bueno, which is good for solopreneur builders.
You don't have to have a big production to do this.
I love that.
I love that.
Maybe one sort of thought here,
and I'm kind of thinking maybe what we should do
is like put together a marketing plan
and kick off a bot to help us take all this advice
and like write us a plan.
So maybe we could do that,
but I'm curious for like maybe a last piece of advice
from you Cody around traditional businesses.
And this is something you've obviously done extensive work on
is sort of like how do you help traditional businesses
and kind of like, you know,
enable those businesses with AI,
kind of help them grow, embrace new technology.
You know, obviously we're thinking about something
that's a pop-up,
which in many ways relates to traditional restaurateurs
or traditional businesses themselves.
And yeah, I'm just curious for your general thoughts
on like tech enabling,
tech enablement for those types of businesses
and maybe the components that you've seen work best
or not work best.
So we're not too over our skis on the whole AI thing,
but yeah, we just love your general advice here.
And maybe we could feed that into our bot
and like come up with a marketing plan.
I don't know.
Yeah, sure.
I think we're working towards that.
That sounds good.
Yeah, I mean, well, let's see.
So we've brought in like 1600 businesses
over the past two years.
Predominantly home services, professional services,
some online businesses as well.
But almost all of them,
they do a ranking in the beginning when they come to us.
Almost all of them are not using AI anywhere near
as intensely as the people who are on here.
They're Googling chat GPT for things.
So the cool part about what you guys are doing
is you're gonna be light years ahead
of most small businesses.
So that's one.
I guess if I was thinking about how to use AI,
I always think Charlie's Munger says like invert, always invert.
I would definitely start from where can you make
small businesses cash?
Because a lot of people are trying to sell them AI services
that don't relate to revenue,
but the average small business multiple is 15%.
So they don't have a lot of cash typically.
So if you can't tie it to an immediate revenue stream,
you're probably not gonna sell your AI services very well.
So that's just one thought.
But if you can do something,
like we're talking right now about how to market the bot
that you're doing and how to spin it up,
I might do something like this, like for instance.
So like what data do you guys have that's unique?
So if you're in this pop-up and you have GROC,
maybe you're pulling together a list of all the users
who are on here who are interested in AI in the area
and you did some sort of survey on them
and you're gonna give it to everybody who signs up
for the restaurant and they get like a free dish
or something when you come.
For ours, like our lead magnet is,
hey, we've analyzed 15,000 small business owners.
We actually used tech.
We didn't use GROC for this one.
This is a mixture of Claude, Vercel and Replet.
But to come up with a survey
that actually delivers something to the user
that makes them feel like they're getting more value
than you are for their email.
This one in particular kind of tells you
what kind of business owner you are, for instance.
So I would think about for you,
instead of just trying to get emails for signups,
how can you give them something that's like so valuable?
They almost, they feel like the value switch is so hard
they wouldn't give it.
They wouldn't feel bad giving you their email.
And then we posted it on X too, so you can see it.
And you start like, I would just do this
with like five or six things.
You could spin up a lead magnet probably in five minutes
and have it be geared to what you know specifically
posted on the internet.
People will want the thing that you're talking about
and then sign up to your restaurant.
And that's what I would do.
Got it.
So to recap it sounds like focus on the revenue dynamic,
just give in the business and think of technology
and data and more as an empowering capability.
Is that maybe a fair summary?
I know there's a lot more there
than my two sentence recap, but.
Yeah, I just, I mean,
you're in business to make money at the end of the day, right?
Like nobody does this for charity.
So I think if you always start with first principle
that you want your business to actually drive revenue,
then you're gonna do the things
that'll keep you alive forever.
And so that's why I'm like, go back to like,
what are your leads?
How are you capturing them?
Can you make your leads money?
Can you give your lead something that's 10x more valuable
than what you're asking for them?
And if you can do that continuously,
it'll be hard for you to lose or be poor.
Yeah, amazing.
Sounds really good.
I mean, this is all stuff
we never would have been able to figure out on ours.
Or we're making progress
and I'm super excited to keep building some of this stuff.
Cody, any parting thoughts for us?
I know you've been incredibly generous with your time
and this has been really helpful.
We're gonna definitely go and throw a bot at the transcript
and get this marketing plan working.
But yeah, any parting thoughts for us
or for the folks listening in
before we kind of switch back into build mode here?
No, I think you guys are doing something really cool.
Just remember, you know,
business is there for you to profit off of it.
Otherwise it's way too fucking hard
and it takes too long and it's pretty miserable.
And the only way that it ends up being fun
is for you to succeed and make revenue and capital.
And so I think that's the only thing
I sometimes see with builders
is they'll build a lot of cool stuff.
But if you build it, I promise you, they will not come.
All right.
Well, the Cody Sanchez,
thank you so much for spending time with us.
This has been incredible, learned a ton.
And obviously we need to go focus
on making money it sounds like.
Yeah, helping other people make money too maybe.
I think so, yeah.
So yeah, thank you so much again.
I think we're gonna take a short break.
We're gonna get some things sort of on our side,
maybe like five minutes.
We'll be back really soon.
And then we're gonna show you guys
what we've been up to, what we're gonna start building.
Yeah, hopefully we'll have the landing page up.
All right. Awesome.
Love it.
See you all soon.
See ya.

Test, test.
We are ready to start.
OK, everyone hear us OK?
I think we are back.
Awesome.
We're back.
We're alive.
Thanks so much for joining in.
What we're going to do is we're going to give you a quick recap.
We've been just hacking on things a little bit here.
We've got about five minutes until our next session starts.
And that is Grock Bot for engineering,
which should be a good one.
But in the meantime, what we did this morning
was we did some market research.
We looked over Twitter and got all of the ideas y'all
suggested for a restaurant.
We aligned on what we're going to build.
We talked to Pierre Yang and also Cody,
who helped us learn a little bit more about go-to-market,
about what we're building.
And then our focus from here on out really
is just going to be building stuff.
So I think we're going to focus mostly
on showing Lauren, showing Potato, doing her thing
after the engineering session and just showing what we can
build.
But the idea we lined up with is a pop-up
for a restaurant, a pop-up experience in San Francisco.
So our goal is going to be to connect basically a restaurant
to a bunch of people that want to eat food
and have an experience.
And through creating that experience,
we'll have some sort of software.
And so the product, ideally at the end of the three days,
will be both an operating system, a product
for creating these sorts of experiences,
but then the experience itself.
And I think that's what we're going to push for, right?
Yeah.
I think that sounds.
It's ambitious, but we'll try our best.
Yeah.
And I wonder if there's a way we can make it some fun too,
like add maybe some, I don't know, we'll play with it.
Maybe we can make the booking experience more
than just a traditional booking experience.
But I'll save the crazy ideas for day two, maybe.
Awesome.
Well, I mean, in the meantime, Lauren,
you said you built some stuff.
Maybe we could start talking through what you got.
Well, I'm still building it.
But essentially, I'm trying to set up my potato factory.
So while we were off stream, I
set up a new bot called Hashbrown.
And Hashbrown's job is to review the PRs that Tater opens.
And as a reminder, Tater is my engineering bot.
And so, hey, Hashbrown's going to be my reviewer bot.
And then it got me thinking, I guess
it'd be nice to have a similar setup for not just myself,
but also Roshan and Matt.
So I'm doing a couple of things now
where I'm adding PStack to the repo.
So one thing you can actually do with cursor plugins
is you can install them both locally on your computer.
But you can also set them up in the repo as our repo plugin.
And your bot or agent can do that very easily for you.
So that's going there.
And then at the same time, I'm
getting my bot to set up a cursor automation.
And if you're not sure what a cursor automation is,
it is a way to, well, think of it as a way
to kind of react to things.
So when I open a pull request, I want the pull request
to be posted to our Slack channel.
And then there'll be an automation that will pick that up
and review the pull request and maybe even merge it for you.
I'm not sure what we'll do yet.
But I hope we can get like this really nice flow where
your bots are just dumping your PRs into that Slack channel.
And then the reviewer bot goes off and reviews them.
And then if it's good, we can merge it.
And then I'll try to make it so that the code base is really
easy to contribute to.
So yeah, a lot of things to do.
There's a lot of scaffolding you're talking about, right?
Yeah, it's a lot of scaffolding.
Yeah, a lot of skills for agents and things like this too.
I think I feel like it's important to set up.
I also don't want to overcook things because we're still
a startup, right?
Our goal is to survive until the next day.
And I think we want something basic, nothing too fancy,
but something we can extend later.
So that's going to be what I'm going to set up
after the 30 minute break or the talk.
Yeah, and I mean this is like the same factory you use
at SpaceX, right?
Like this is very similar to the process.
Except this time I have to start over, right?
I don't have all the stuff that I built before.
But it's a good chance for me to maybe even rethink
how I did the original setup.
Maybe there's a better way, right?
Maybe I'll get inspired as we build out this new factory.
100%.
So we're going to take it back there, I think.
And something we talked about during the break also
is that I might take the lander.
So I'm going to think through visual design.
And typically I build with PSTAC.
And my flow is like using GrapBot to ID8,
pushing that to cursor cloud agents
to actually build this stuff.
But we got about one minute here.
And then we're going to kick it over to GrapBot for Engineering,
live from San Francisco at Dreamforce.
And we're going to have one of our teammates
walk through exactly how you can use some of these workflows
that we're talking about live on your own.
So I'll kick it over there.
And we're going to cut there.
When you come back, we're going to just keep building.
So I'm excited to see you all this afternoon.
Good afternoon and welcome to GrapBot Galaxy.
And thank you so much for joining today's workshop
on GrapBot for Engineers.
My name is Ling Shi.
And I'm a software engineer at SpaceX AI.
To give you a little background about me,
I joined Cursor, which is now SpaceX AI right now,
for nine months, and already shipped two products.
I previously built Cursor 3 Asian Window
and now building GrapBot.
And interestingly, in the past two months,
I've been only using GrapBot for my engineering tasks.
And with GrapBot, I was able to build the first version
of the GrapBot mobile by myself in three weeks.
And I can also learn much faster now thanks to GrapBot
because I can get to work on a lot of different aspects
of the GrapBot and also apply different approaches
that I think that we didn't try following the first principles.
And the more I build with GrapBot,
the more I want to share this superpower with you all.
And this is what today's workshop is all about.
So our whole engineering journey
started at around from the early 2024.
At the time or before that, we only
have like autocomplete, which is like syntax-based.
And after that, we started to have tap completion,
which is already a big leap forward.
And since then, we started to have the ask and edit.
That is the first time we can feel the power of the agent
where it can help us glimpse through the co-base much faster,
much easier, and also start doing some repetitive tasks.
But at that time, they cannot do a big bunch of tasks.
And we just keep iterating.
And then the agenting coding comes.
The cursor 2 and cursor 3 are all about agenting coding.
And it is able to manipulate your computers,
manipulate your mouse, your keyboard
to test things end to end.
And not only that, it can now sink on a longer horizon
to know what needs to be done for the next step.
And it can also plan things.
It can work on things like we have the slash goal,
slash loop that has been pretty useful on getting agents finish
a task end to end.
So I don't have to nudge it.
But at the time, we still need to nudge it.
We still need to write the prompt.
We still need to start the agent from some sort
of interface.
So the next step comes, the automations,
which is running on the cursor Cloud Agents.
That is a different era because the agent could then
be started based on some triggers,
like you receive a Slack message.
You get something.
And that unlocks a lot of possibilities
because it can not only rely on you starting the agent.
The agent can start on its own that you are not
with your computer.
The agent can also run 24-7 even when you are asleep.
It can also start based on the Slack messages it triggered.
And another thing that it unlocks
is it no longer limited by how much your local computer can
run.
The whole idea of Cursor Cloud Agents
is that it can run whatever is possible on your computer
on cloud.
That means we can just give you 10 machines or even more
to run things in parallel.
Your agents will no longer fight for the same port.
And at the time, the Cursor Cloud Agents
has unlocked everyone.
Everyone at Cursor Team has almost
been 10xed or even more.
But we have also been thinking through,
what should be the next step?
Like, agent coding is already very good.
We have seen promising results that a small team can
ship a big product.
What's more?
What can make our life easier?
So the idea is we can possibly make Gropbot,
we can possibly make some AI agent to help us
write the prompt, to help us orchestrate.
Because before that, I was able to manage 15 Cursor Cloud
Agents.
And it was already a lot.
I have been context switching between different Cloud
Agents.
But I really wish there was another agent that
can help me kick off the follow up,
help me queue the messages, help me nudge or interrupt
the Cloud Agents when I think it shouldn't continue
or it needs to be steered into a different direction.
And we do believe that the models nowadays
have the capability to get the signals we need.
So then we just put them together.
And now it becomes the team of autonomous agents
that can act as your colleague.
And you can directly tag them.
You can directly work with them.
And they can grow with you.
And that is Gropbot.
So let me introduce you to the Gropbot.
This is a team of fully autonomous AI agents designed
for all types of works, including engineering.
First of all, it runs 24-7.
And it is fully autonomous.
It can run on its own.
It is not limited to when your computer is asleep or not.
And it also does not require you to actually have a laptop.
The second thing is, which is the biggest unblock
for the engineers, is it can manage coding agents for you.
Not only cursor-cli agents, it can almost
manage any type of coding agents for you.
And the best thing is it has to f-

first-party integration with cursor cloud agents.
So it has the tools to be able to read the transcripts,
to start a cursor cloud agents.
You can also start on a private worker
so that if you have your Mac machines you want to learn,
they can start into the Mac machine.
Anything you can do in the cursor window with cloud agents,
GragBot can do for you with tool calls.
And so you don't have to let your GragBot control
some user interface to start the agents
and read the transcript.
It can efficiently read all these through their tool calls.
And to me, this unblocks a lot of workflows
because it can examine what's being done by the cloud agents,
not only being done but also what's in progress.
So for example, I have some instructions to my GragBot
that after the workflow is done,
it can check what's being proved in that workflow
or in that in that cloud agent term.
And then it can use my instructions to determine
whether that is treated as completed
or it's missing some important proofs,
for example, like screenshots or the performance comparison
before versus after.
If it's missing, my GragBot can just directly create
a follow-up reply to the cloud agents
with what needs to be done so the cloud agents can then
continue working even if I am not with my computer.
The second thing is it can connect to most of the tools
you use.
For example, if you are an engineer deploying
your websites or dashboards to the Vercel,
it has the Vercel MCP.
It can connect all these deployments.
So it's not like connect a Vercel MCP
with your agent take coding tools.
It's that it can just run 24-7 and it can check the signals
when the build is failed, starts to work on that.
It can also directly kick off a deployment
based on some timing.
For example, you want to get a deployment on 6 AM tomorrow
but you are a night owl.
You can just ask it, say, OK, deploy 6 AM tomorrow.
It will know the timing.
It will call all the tools you use at the time.
So you are no longer needs to start a local agent
or some other agent at the time you need them.
Another big thing is it has the memories and routines.
This is the biggest on block.
This is actually a big leap forward
comparing with what we had before with continuous learning.
It can persist the memory based on what you tell them
and over a very, very long horizon.
And also the Grogbots are named.
So you can have different memories
under different Grogbots so that they are not overwhelmed
by tons of memories that are possibly unrelated to what
they are currently working on.
And this is very important because most of the time
when you are talking with Grogbots on something
that they should be doing better, they can just
remember it.
And the next time when they start a Cloud Agents
or doing some revision works against the Cloud Agents,
they can apply what you previously enjoyed
and then directly use that into the prompts
that they sent to the Cloud Agents.
You don't have to repeat what you said
thanks to this memory system.
And I think it's also important to understand why you need
a Grogbot because there are a lot of products out there
that are possibly able to do the work that you might
think they're able to do.
For example, I was a big user of the OpenCloud and Hermes
agent and they were not really able to do the engineering
works that good because they lack the first party integration
with the Grogbot.
So I think the first idea is you don't need any caffeinated
laptop anymore.
I think that's a big on block because I know a lot of you
maybe have a tool or some terminal commands
to make sure your laptop is awake just
to get the things done.
We don't think that's an elegant experience
for any sort of engineering tasks.
We want to get rid of it so Grogbot has
to your own computer 24-7.
Your laptop can stay awake, can stay away from being awake,
and can just keep being a terminal for sending the commands.
And after you send it, you just close the laptop.
You don't really have to worry about if anything
going to stop working or stop running.
The second thing is it's beautifully available
on all platforms.
We recently shipped the iPad and Android
and it's already available on iOS, Windows, Linux, and macOS.
So you can literally work on anywhere.
For example, you can lie on the grass on the Golden Gate
Park while your body's working.
And the third part is also one of the biggest on block
for the Grogbots, comparing with other products,
is that you can control the computer when needed.
Unlike other products, when they actually
need a login or they actually need someone to help them
click through something, you cannot do it.
You can only prompt them and helplessly
see what's going on on the transcript.
It's no longer a case for Grogbot.
On any platform like iOS, Android, or iPad,
you can directly click through the computer
to get things done, whether it's
like log into some services that are very hard to navigate
or require some sort of human verification
or when it's doing some login but you don't want to hand
your pass right away.
You can use the mobile clients or our desktop clients
to click through the remote computer to get things done.
You cannot really get this with other products.
And the last part is the first part
of coding agent integration.
This is also only exist at Cursor
that our Cursor Cloud agent integration with Grogbot
is very, very smooth that almost everything
you can do in the user interface of the Cursor,
it's possible with Grogbot.
And let me share a few interesting engineering use cases
that I really enjoy.
It's been empowering me since the past two months
and I really love what we have been building around Grogbot
and also build a Grogbot for us.
The first thing is get the things done when I'm away.
So let's think about you are possibly asleep
or you are on a flight and your teammate needs an unblock.
You are the only engineer having the code
on or off of some service.
Previously, they have to wait until you
get off the flight, which is kind of sad
because in the modern startups, the speed
is very important.
So let's think about you are getting pinged
and your bot can monitor a slack
and based on the situation or the criteria as you said,
automatically unblock them, reviewed for whatever is needed,
and then send an acknowledge back in the slack.
That's what's happening right now.
And it's been pretty smooth because I
don't have to be always next to my laptop,
even if something that's actually
required my review or required my supervision.
I can also take a look on my phone.
And then once I feel it's good, I can just
send this through my Grogbot saying, OK, this looks good.
Approve it.
And then the Grogbot will just send approval
and also notify the sender about this thing
as being approved.
They can continue.
And also, we can possibly think further about what
if they can monitor just the general slack for me,
but not only when I send them something that says, OK,
it's approved.
I can tell the Grogbot that these are my review criterias.
I need something to be attached with a screenshot.
I need it to be attached with a test that are not fake tests.
And then whenever someone send me or attack me on a slack,
they can read the message, check if it's a review case.
And then they can kick off my skill
for doing a code scan and also doing code research based
on the criteria as I gave.
They can also run it in a cloud agent
so they can execute a skill inside your repository.
And then they can acknowledge it back.
So I don't have to be awake like a side of my computer
or even a side of my phone at all
and only be not when my bot actually needs me.
And how about one step further?
How about monitor X from?
There are a lot of bug reports about the Grogbots on X.
And those are very valuable feedbacks.
So we also have set up the bots to monitor the slack
thanks to the first-party integration with Slack MCP.
So we have the bots to automatically kick off a task
to start digging into whether the product is still valid
on main and then if it's still valid, go fix it
and send me the PR.
And the same thing applies to other surfaces
where we need security reviews or we
need to go find the right person for them to review.
Everything now is bots contacting with bots
and the human only connects with human
when it's absolutely needed so we can focus on a deeper work.
So that you might be interesting in where
is the human stands in if bots can do everything for us?
That's a very good question because humans still
have a lot of things to focus on.
I still drive the product direction.
There are still a lot of design details
I need to dive into.
There are also a lot of difficult things
that bot couldn't finish like performance issues,
like how we should architect our next version of the app
so we can avoid most of the issues we have
and also have a good architecture in the front
so we won't get sloppy code along the horizon.
I also need to give them a better instruction
on how to say no.
I think this is a very important aspect in the product
because there are a lot of bug reports and feedbacks
and some of them is not suitable with what
you need in a product.
The bot is very kind.
They will possibly accept everything by default
so you need to set the boundary.
You need to tell them why you need to do this,
what kind of things should not be doing
and here's the reasoning.
The bot has the memory system.
They can remember what you told them
and then apply it to the future decisions
so you don't have to repeat, say,
I don't want this in the product.
This is not simple enough.
We need to simplify this.
You don't have to repeat that over the time.
And also, I need to make sure my bot is not blocked.
I think it's also very important because there
are a lot of authentication steps in between.
So the GROC bot needs to have a very good experience
on getting the bot unblocked by a human.
The next use case is also one of my favorite use case,
which is available on the marketplace as well.
It's nightly code cleanup.
We all know that in the past year,
we have been possibly shipping some sloppy code.
Everyone hates that.
And sometimes it's just hard to enforce every single agent
to apply the same rules.
So another solution is just to clean those slops out
over the night.
That's when the people are not shipping,
so it's less likely to hit a conflict.
And it's also less risky because most of the time,
they're just slops.
They can just be condensed into a smaller module
or remove the lengthy comments that you just hate.
Those are very low risk.
Could be done over the night.
And you can also trust the cursor cloud agents
to merge those PRs based on the conditions you set.
For example, you have provided a clear end-to-end proof
of what's going on.
If that's provided, you can feel free to merge.
And right now, my setup is every 3 AM, my bot
will start a research cloud agent
to review across the whole surface of our monorepo.
And they can identify issues like code quality,
not being good.
Things should be modularized but not properly modularized.
The comments that are not properly condensed
into what is necessary.
And also security audits that sometimes we just
forgot that this is very important to prevent someone
else to stow something.
We have to grab our catch all these kind of things
over the night.
And when I wake up, I just have a set of PRs available.
The next thing is about internal tooling
that this is also changing how we collaborate
within the organization.
Previously, for things like adding people
to the test fly via email address rather than via a link,
you might think, I can build an internal dashboard,
put some authentication on it, put some database against it,
deploy to a Excel behind some authentication wall,
and then distribute it to the internal employees
to sign up for the test fly.
It works.
And with the agent coding, it should be already not so.
You can get it done by in an hour or two.
The thing is, why do you have to?
With Grogbot, one of the flexibilities
is that Grogbot can run logics with a much simpler
interface, for example, like a Slack message.
So the unblock here is that I have everyone just
mentioned my bot inside Slack and just put the email address
next to it.
My bot will have a routine or some web hook
signals based on Slack.
So whenever it's one tag, it will
start to run the routine of adding people into the test
fly.
So the only thing I need to do is
to make sure my bot has access to the test fly
in a safe way.
And other than that, I only give it
one prompt of what needs to be done,
and this internal tooling is available to be used.
The next thing is also my other favorite use case, which
is Out of Fix Everything.
You might have a lot of deployment pipelines, or CI's,
or end-to-end integration tests.
The biggest thing is when we actually have a CI going red,
we usually ping on call rather than another bot
to get it fixed.
That's changed.
The on calls are only being involved when it's absolutely
needed.
And most of the time when CI is red,
a bot will automatically kicks in,
starts to examine what is being read,
and then starts to spin up a Cloud agent to fix it,
and then also has our instructions
of how to determine if you should merge it
and then they can just merge.
So the turnaround of needing a human to take a look
and the human needs to contact switch between tasks
are very, very minimal right now.
Not only for a CI being main, any deployment errors,
any CI flakiness, or any sort of tasks that has a signal
or has an alert currently in your internal pipeline
could be hooked up into the Grop Bot and only ping on call
if it wasn't resolved after like 10 minutes.
Most of the time, Grop Bot can get it done within 10 minutes.
And right now, let me introduce you
to how I organize my bots.
So we are kind of in between a world where you only
need one bot versus you need a fleet of the bots.
So I have my chief of staff, which is called Ling Xixi.
I have the Craig working on the UI engineering.
I have Steve working on the DevX and Hogan
working on Infra.
You might ask this question of why we need three engineer bots
rather than one because apparently they
are sitting on the same model and they
might be able to do the task.
100%, they will be able to do any task you gave them.
Even if I gave a UI task to Hogan,
they should be able to resolve it.
The issue is the context and also how
they're going to manage their fleet.
Every single bot has a slightly different pipeline
based on the work they do and also
has their own context limit.
To avoid they run over their context limit
when switching between too many tasks,
I prefer to put them separately.
So whatever is told to the Hogan will
be persisted in the Hogan's memory.
So all the instructions, et cetera.
The next time when a new task come in,
the Hogan can just start the new task based on the context
it already has rather than trying to pull the instructions
again and then start the task in a new area.
This can make your fleet runs much faster.
And the key thing is you don't really
have to talk with your individual engineers unless necessary.
You can mostly talk with your chief of staff
and the chief of staff can route the request
to every individual engineers because chief of staff
has the understanding of who is working or what.
While chief of staff can also save his context
from remembering all these engineering workflows
or how things should be developed.
And now let me give you a demo of how
I can set these things up.
This will be a very life demo because I'm also
the first time setting these things.
So right now I just downloaded Lin Xi's engineer
bot from the marketplace.
And it's onboarded into what kind of repository
it needs to be done also connect to the notion.
Right now I'm going to download the nightly audit
bots from the marketplace.
So you can find it from the GROCK bot team right here.
We all.
There's like a nightly audit engineer.
I will edit.
So now it's very simple.
It's already started doing all these onboarding tasks.
And the thing is this nightly audit engineer
does not have the context of how my existing engineering
workflows is being done.
So what I would do is instead of telling it all these things
that I will possibly already told,
like follow this pipeline, do this first, do that then,
I will have my existing engineer bot
to onboard my new engineer.
Hello.
I have a new member in the team called Nightly.
Rename them to Steve and tell them
how the engineering workflows are enforced.
So my engineer bots, actually I should rename them to Craig
because that's my name for them.
My Craig will send a message to my nightly audit engineer,
which will soon be Steve, and then
starts to handshake all these things over.
So you can see my Craig just messaged the Steve
with some message of how everything goes.
It includes all the requirements of how our notion board is made,
what is the definition of the clean,
and how to ladder through different stage of my workflows.
All via plain text.
And my Steve will acknowledge and absorb into their memory.
You don't have to repeat or copy, paste
every single thing between bots.
You just let the bot talk with each other.
You can see they also just confirmed between each other.
So I don't really have to talk with Steve.
Like my Craig will just tell me everything is confirmed.
He's all good.
And right now, the Steve is equipped with the nightly
audit skills.
So I'm going to kick this off right now just
for demo purposes.
Usually, it runs on 4 AM.
But let Steve know to start a nightly audit workflow right
now.
I just let him stop right now.
So in the meantime, we will be waiting for the Steve
to do some research and get back
to us with all the findings about the FlyLo.
We will take a look over some other examples
of how I can work with the Asian in parallel.
So right now, I already had a task to Craig.
And I think there's one issue with the FlyLo website.
That's being reported.
It's very urgent.
So OK.
I think there is an issue with FlyLo.com
that you should cannot check their previously booked flights.
Can you check if it's true?
So my bot will be able to navigate a website
in their computer, click through, and show me what's going on.
This is one of the unblocked because we
know a lot of engineers has workflows that are not
reachable through MCP.
So if you need to click a button based on some instructions
or at some certain timings, you can ask the bot
to do it by using their computers.
So it's checking the FlyLo.
I can monitor it from the computer.
Yes.
So it's currently navigating through the website.
Try to understand whether this is a legit issue.
So it's navigating through.
You can see it can generally do anything you need on a web.
So almost all of your workflows could
be delegated and unblocked.
And I think because this issue is
like being reported by the user, I want it urgently.
But the thing is when you talk with agent,
they possibly know, OK, the urgent means
I need to get an ASAP.
But what it might do is ended up skipping the steps
or start guessing things to actually get
sure it results faster.
We don't want that.
We don't want agent to guess the same.
We want deterministic results.
So previously, before Gragpa, I may tell them to just repeat
them like I need the urgent clay, for example,
like I need the urgent clay, and tell them what you need
to do and what you should not do.
But how about extract them?
I don't want to repeat myself every single time.
That is the whole point of the Gragpa or the autonomous coding
that you should not repeat yourself.
And I will let my chief of staff onboard a new member called
Jenny, onboard a new member called Jenny
for doing all the head of operations.
And my Jenny will start to play around with the playbook.
Jenny needs to manage a playbook for my engineering team
and needs to on the playbook and avoid other engineers
to update.
So my chief of staff will start to onboard Jenny.
And then my Jenny will start to on a playbook in Notion.
And then they can update that playbook based on what I tell
them on what needs to be done at what time so that every other
engineers can just apply the same playbook.
You can see the Jenny is up in the stage.
And my chief of staff is also doing agent to agent
chats between them so that they can just handshake
to each other.
I don't have to actually tell Jenny when it's to be done.
The only thing I need to talk is with my chief of staff.
Let me actually check in with Steve on how much it's going.
OK, so the Steve, without my instructions,
oh, this is my built-in instructions.
So without any of my chat, Steve is already
doing the research on what kind of things
could be improved over the code base.
And it has already started a bunch of the Cloud agents.
And it seems like they are finished.
So my Steve is currently working on and conducting them
into the PR that actually improves them.
And then now let me get back to Craig to see what is going on.
OK, so he says it's confirmed.
The issue is true.
SlashTrips is a stub.
While the navigation still links there,
there's no check today for a previously booked flight.
So now it needs to do this urgently.
I really wanted to do urgently.
But let's slow down a little bit right now
to think what they need to do when we say urgently.
There's, I think it's a widely acknowledged issue
that coding agents sometimes are running very, very long
horizon, and it's not good when you need things urgently.
So one of the idea to help unblock this situation
is to have you checking their syncing traces,
checking their tool codes to make sure they're on the right track
and not to do them or steer them at the right time
when they are possibly going off track.
But that requires you to be next to the desktop.
And sometimes you have other tasks to do.
You're not next to your desktop.
And you don't want to look into the details, syncing
traces, et cetera, every single second.
So what I usually do is I have a P0 escalation workflow
with my bots that let me define this.
So treat as P0.
And the P0 definition is they would need to monitor.
The GROC bot needs to monitor the cloud agents every five
minutes to check the progress, to check where the cloud
agent is currently running on.
And sometimes the cloud agents are running tests for a very long.
They do like a sleep 300 to just get around the tests that
are running async in another shell.
And sometimes the tests are fast, like within a minute.
Then you've got four minutes lost.
So when the GROC bot can check into the cloud agents,
they can check if they're doing a tool
code that's very unnecessary, time consuming.
And they can interrupt it immediately
and start a new prompt based on what's happening.
And it does not need me involved at all.
I don't write any prompt at all as well.
The only thing I write is this.
It's the definition of the P0.
You need to set up a routine that checks cloud agent
every five minutes.
Check if they are off the track, such as running a long sleep,
like sleep 300, or going off our goal, being too conservative.
We sometimes just need things moving a little faster.
And this is a very rough prompt I give it.
It can definitely grow along with your interaction
with the GROC bot.
For example, you see it possibly work better.
In some situation, you can then just
tell it to do that.
You don't have to write it into a skill anymore.
You don't have to write a long prompt every single time.
You just tell them once in a normal message
as how you would tell your teammate
for doing something urgently.
Going off our goal being too conservative
and not interrupt and notch them at a time,
you found them going off.

because I need this urgently, urgent, urgent, urgent.
So I'll also tell them to fix the issue.
I can just click reply so that it has the context
of what I'm quoting on.
Fix this issue urgently.
Treat it as a P0 and then it knows what is a P0 now.
But only Craig knows what his P0 is.
Rather than copy paste into every single one
of my engineer bots because that won't scale
if you have 10 or 15 or 20 bots.
I will have a handed with the Jenny.
So let me check if Jenny is on right now.
Jenny owns edit, Jenny owns edit.
So Jenny will nudge me whenever they need a decision
and I think that's a very good question for me.
And they just managed message the bot,
I think everything is set up.
And now I will let the Craig to talk with Jenny.
About this P0 workflow and bake it into the playbook.
So now my Craig can then talk with the Jenny
to put this into the playbook
and then Jenny can go announce
to other engineers automatically.
And then also let Jenny to announce
to every engineer bot.
The good thing of this is you are just thinking
what needs to be done once.
So you finish the workflow, you get the handshake on.
So all your bots understand what needs to be done
at what time.
The next time something happens,
or the next time you need to add a new workflow,
like you actually need to do this,
you need to do some step
before you actually trigger deployment.
They can just populate the playbook
and all engineers will know that
without you telling them individually.
This is very similar to how a human organization is shaped
while it's at the agent's scale
that Asian has a context.
Sometimes human does.
So the treatments for human teams
sometimes apply to the bot.
Some does not apply,
but you can always experiment
whatever workflows you have and see if it works.
Okay, so they told the Jenny,
you can see most of the conversations
are directly between the bots,
just like I'm managing them as an engineering manager
or I'm mentoring my talented intern.
So they kind of know
what kind of things needs to be handshake between them.
They just don
the bots.
That's also how I work nowadays,
that I don't really check my cloud agents.
I check the results and check the proof,
but it will be very disappointed
if the proof is just not included.
So I have to nudge them back,
say, oh, this misses the proof.
That is very bad.
I should just do it once
so that all the things onward
are just included, proof automatically.
So I'll just tell with my Craig,
which is kind of like my head of engineering right now,
that another requirement for the playbook
is to make sure every single
PR comes with proofs like screenshots
for UI changes and perf metrics
for performance improvements.
They should not understand
what to talk with the Jenny.
And let's see how it goes with Steve.
So Steve already gets something back.
You can see some cleanup and the PR is ready.
I can directly click over and start the review.
And another one is also still running.
The best thing of the Grog Ball is,
even if you are chatting with them,
they can still do their existing things pretty well,
just like a human who can multitask a little.
Though sometimes if you nudge them too hard,
they will start to lose their contacts
because there are too many tasks to handle.
So that's where my original engineer bot comes in,
which is you can get from the marketplace,
that it comes with a fleet notion database
that defines every single task plus the stage they're on.
So even if one bot is managing like 20 cloud agents,
they don't have to populate their limited contacts
with everything about all these 20 cloud agents.
They can just always check back to this DB
whenever they need to take the next task.
So they don't really have to remember
what is currently going on,
what kind of cloud agents are being working on,
and what is done and needs their eyes for the next steps.
You can see this is a trust polish,
which is watching 3.3,
which is going through the watching process
of what possible security comments and bug bots.
This is also a big unblock in the engineering pipeline
where the bug bots can help you identify issues
made by your agent and security common feature
can also monitor potential security issues
with respect to leaks or session management, et cetera.
So that's very helpful.
So usually in my workflow,
I would just go into this PR and review it
and then get emerged.
Let me actually check this PR.
So this is that cloud agent
that I previously kicked off before the stage
and it got some screenshots of what is being done.
Plus you can see my agents are now writing all the prompts
for my cloud agents.
I no longer have to write all these prompts though I could,
which is what I previously sometimes did.
I don't have to.
The agents would do this for me
so most of my tasks can stay on
what kind of things needs to be done,
what kind of things needs to be killed.
And you can see this task just gets back all these screenshots.
I can just feel confident to merge and what I will get.
Okay, so go back to the presentation.
I think we have,
I have been sharing a lot of my workflows today
and I wanna wrap up on what we can learn from today.
So you can take away with some ideas
on how to start this whole workflow
and how you can leverage it.
The first thing I think is very mind blowing
is treat them like interns.
So when you feel struggling about like communicating
with robots you don't know what kind of things to say
or they just doing bad.
Don't treat them like an agent
or try to invoke skills, et cetera.
With Scrockbot you can chat with them in a messaging base
which is how you would chat with a talented intern
who does not know about what to do
but once you tell them they can be extra capable of.
Another thing is we can possibly think one level further
rather than just fix the thing.
This is also could be known as reduced repetition
that sometimes I repeat a prompt to my agents too often
that it's a waste of my time
if I have to type in or copy paste the same thing
and send it to the agent 10 times.
You can think it one level further
about what actually needs to be done
and then extract it into some layer
either via a playbook or via some other bots
to nudge another bot for you at the right time.
The last thing I think is also very important
for using engineering with cursor cloud agents
is to have a complete feedback loop.
No matter what type of engineering tasks you have
it needs to have a feedback loop in order for agent
to understand the signal of success versus failure.
That's also very helpful for Grogba to determine
whether they should nudge further
or whether they should stop and let it merge.
For example, you can let your agent
to control the website via computer use
and I think that's the easiest feedback loop.
For hardware engineering there are also feedback loops
on the signals of whether something is actually good
or not.
And thank you.
That's it for my workshop.
Thank you so much for your attention.
To those of you joined virtually,
it's time to head back with Matt, Lauren and Roshan.
What we're gonna build
and maybe a good place to start is that Lauren
could walk through some of the things
that she started shipping.
What do you got Lauren?
I haven't been doing anything that interesting
but I am, so the background is in the SpaceX AI Slack.
We have a lot of automations that do things like
review our PRs, automatically triage bug reports
and stuff like that.
So I wanted to get something basic set up for us
so that your bots can post PRs to the Slack channel
and get it reviewed
and then you can merge it
and we can figure out how to deploy, auto deploy maybe.
But yeah, a lot of the ops stuff I've been trying to work on
and I can walk through what we've had so far.
I don't know if my screen's being shared but
are we sure?
Yeah, we'll get Lauren's screen up here
and we'll start talking through it.
Yeah.
Maybe while Lauren is doing that,
we can just recap,
while we're waiting for the screen I should say,
we could recap a little bit of what we're building.
Mr. CEO.
All right, so what we're working on is a restaurant pop-up.
So we're gonna do a pop-up in person in San Francisco.
This is a bit ambitious.
So our goal is gonna be to collect
both people that wanna attend the pop-up
and then people can actually make the pop-up happen
because while we're building the business,
we're not restaurant tours.
So we might need to find a restaurant,
we might need to connect people,
find a location, do all the logistical work
and then we're gonna build a platform around that.
So the software that we build to make that pop-up happen,
we're gonna need to figure out a way
to turn that into a product and to sell.
So it's a bit of a stretch goal by day three
to make some money
but we're gonna do our best to make it happen.
And this is my screen where I'm connecting GitHub
and setting up repositories, so shipped by Thursday.
But if we could get over to Lauren's screen,
who has more things to show you,
this is a permissioning for cursor cloud agents.
And so we'll work on getting over to Lauren's screen.
But in the meantime, Lauren's working on building out
her sort of software factory.
I'm connecting some GitHub repos to do some design work.
So typically my workflows start in Grockbot
and then they move over to cursor.
You can kind of see how Grockbot's interacting
with cursor here.
But yes, once we get Lauren's screen up,
we should be able to start talking through
how she uses Grockbot.
But in the meantime,
we're working on building a restaurant business.
I'm acting as CEO, not self-appointed CEO.
Somebody else said that, so throw that out there.
Lauren gave us all titles.
Lauren gave us all titles.
Rishan is the CEO, Chief Product Officer.
Lauren is the other CEO, Chief Potato Officer
and also the CTO, Chief Engineer, building stuff.
So yeah, until then,
I'm working on getting cursor set up here
to add some repositories.
So typically you just have to auth your org.
We're using our like shipped by Thursday org here.
So we've configured that.
Also like some repositories that we're building in here.
So we have Thursday.
We have our company.
There's some other stuff we're working with here,
but these are the two main ones.
So I have a domain.
We should have access to both of these domains.
And should be off.
See if Grockbot's good.
On the meantime, I've been creating some bots here.
So we have our like prioritizer bot.
I have my like operator research.
So what I was trying to do here,
we talked to Cody about this,
is that we wanted to understand some similar businesses
that have done popups like this.
So we have restaurateurs, operators
and like sort of a list of these people
where I can kind of see, do some basic research.
And these are all really good restaurants by the way.
Flower and Water, Home Acosta Restaurant Group.
And so now we can like kind of start to understand
the business plan,
but more than that, we're building up software.
We're gonna start to connect people.
And to create these bots, I'm using Dr. Egbot,
which is Lauren's bot you can find in the marketplace.
One of the first featured bots for making more bots.
So about a bit of a bot factory, if you will.
And so now on the AV side,
if we can cut over to Lauren potentially,
and away from me to another screen that's being shared.
Are you sharing your screen, Lauren?
I am.
Cool.
So we'll find out how to make that happen.
And there we go, look at that.
It's hard to share your screen.
So yeah, like I was saying,
now we have a few new channels in our Slack.
We have a PR review channel.
Let me make this bigger.
So you can see that my idea basically is
me or my bots and I will post their PR links here.
And then you can see that cursor has an automation
that we just set up.
It's pretty onto the physically right now.
It's just a prompt in the automation.
Maybe I can actually show it to you.
But here is the automation.
And yeah, this automation is going to review
pull requests for us.
It's gonna use P-Stack.
It's gonna do things like check for correctness,
correctness, risk, missing tests.
It's got a bunch of stuff in here that
I don't know if I like it, but we'll kind of iterate.
And so that's gonna fire for every post in this channel.
And then I'm setting up a new bug reports channel
where I'm gonna set up a very similar automation.
And maybe we'll even figure out like once the page goes live
a way to get feedback from like people on X automatically
or something and pipe them all into this channel.
And then we can set up a little like factory going
where we automatically open PRs and triage bugs
that people report and have that fixed.
Obviously we don't really have anything to report bugs,
but we don't really have much of an app right now.
But I wanna just set up something really basic
just so we can get started.
So.
Cool, well, awesome.
The other thing I'm working on is this thing.
So I talk a lot about this on Twitter or X,
but I think verification is a really important part
of using bots well.
And I'm not sure why the screen is frozen on Slack,
but verification is really important
because it allows your bot to run the application
to take traces to take heat snapshots,
the bug bugs.
And it's really important so that the agent
can do work of its own that's not bottlenecked by me
because without a verification skill,
then my bot will constantly be saying,
okay, now you can run the app and click around
and tell me if it works or not.
And that takes a lot of back and forth, very, very slow.
So what I wanna do is to set up that verification skill.
So Steve, well, I already prompted Steve
that let's first use P stack to run this skill
called create verification skill.
So it delegated to our tater bot.
And let's see what tater is up to.
So I guess it's trying to work through that.
Let's say, let's try, what's the status tater?
And we'll see what it comes back with.
So you might be seeing my screen right now.
I'm just working on some of the lander stuff.
So while Lauren's working through that,
we're gonna figure out how to get her screen back up
while I'm working on the lander.
Yeah?
Do you have any ideas what you're gonna do
for the landing page?
So actually we kind of started wiring something up.
It doesn't seem like,
seems like we're running into some more technical
difficulties where the screen is frozen.
So anyway, we have the lander wired up.
We have a domain that we're prototyping.
We'll share that with y'all shortly.
But in the meantime, we are working on
just polishing the design.
So I'm gonna start kicking off some bots
like work on the design.
I have some skills I use.
So like make design, make interfaces feel better
is a good one.
I prefer minimalism and that's maybe a good place to start
before we start working on some branding.
So I'm gonna take that prototype you built, Lauren.
And I'm gonna turn that into something
that I think looks nice
and maybe a little bit more professional
while you kind of work on the backend services.
And then we can tie in some like wait list
sign up type stuff.
So now we have what appears to be Lauren's screen.
Yeah, you can keep walking through.
Cool.
Is it working?
Yeah, so it spawned a cloud agent
and I guess we're just waiting for it
to build the skill for me.
PSAC plugin.
I guess while we wait,
we can also just take a quick look at PSAC
and what that skill does.
So I think there's like two interesting parts
to verification skills
that in my experience have
makes the skill more useful for agents.
And that is that I feel like there's largely two parts
to the skill, which is one is you want like a CLI
or a script or some kind of
tool that allows the bots or agents to reliably
and deterministically interact with your application.
I mean, bots and agents are pretty smart now
where they can just write scripts on the fly.
But that obviously uses a lot of tokens.
And it's also like, it's not reproducible
because now every bot has its own script
that it's gonna use for doing the verification.
And I'd rather it just have like a standard set of tools.
So the verification skill is going to
or the create verification skill
is gonna create this CLI for us.
And also it's gonna create something called a feature map
which is not super important for us right now
because we don't really have any features.
But the idea of the feature map is it gives
as the names suggests,
it's a map for the agents and the bots to
what do you call it?
Understand the application,
like what features you have,
how to get to it,
keyboard shortcuts and so on.
So let me see why this is taking so long.
Cool.
And I'm thinking like,
I'm working on the wait list right now.
We really just need a way for people to submit emails
and then log that somewhere.
There are a couple of different platforms.
I actually like loops for this
which is actually a really great platform.
You can also use resend.
I might just start with resend
because it's pretty simple to get started
and agents are quite good there.
So the lander.
We might have a plugin for it, I wonder if we do.
Ooh, I think you're right, I think we do.
Yeah, I think we do.
Cool.
So I'm directing my agent to work on that Thursday repo.
We do have a resend plugin, so that's pretty cool.
Yeah, I just have to wait for this.
By the way, the domain is live.
Yeah.
It's connected.
I noticed that.
It's now connected.
Do we have any submissions yet?
Well, we haven't heard of the submission form.
No, I did.
Oh, you did?
I hooked it up to our database.
Oh, really?
So we might, if there's anyone hitting the page,
I think we'll start seeing entries.
Let me see what we got.
Let me see if we got any.
Yeah, I'm going to.
Is this in the back end?
I don't need to go to?
Yeah, you need to go to the planet scale,
and then you can look at part of it.
So that's a good thing to mention.
We already have a submission.
I forgot to mention we have a database hooked up.
Wow.
I think we want to redesign the form first, though,
because I think right now it's kind of vibe coded,
the form.
I don't know if we're even collecting information
that's that useful.
So if you're submitting, maybe don't submit yet,
because it's just a throwaway prototype for now.
But I think we'll want to iterate and think about what
information we need, who the audience is, and.
I have my founding Engbot open right here.
I'm going to do a little rant with what I think we need.
So we need to redesign the UI to be much more minimalist,
first of all, if it looks really bad.
And then second, we want to orient and around collecting
guests or reservations.
So what I'm thinking is we just want an email sign up
for people that want to join our pop up.
And once the person signs up, we're
going to log them in the database.
In the top right corner, we want a button
for restaurant tours or other proprietors or people who
can help us make this happen.
So we're going to collect primarily guest booking
information so you'll be able to go sign up.
But then secondarily, people that
want to submit basically a request to help us out,
because we need some help.
We're going to need to go find some people.
And maybe people want to submit that.
I think we're just going to have to send a lot of emails
and use our GROC bot to do market research on how
to make that happen.
And so I'm going to work on that personally,
but I want this bot to redesign the website.
So with that transcription, hopefully my engineering bot
will start working towards getting that done.
So while we wait for the verification skill,
we'll use AskTutor to use CloudAgent, I guess,
to prototype some admin.
So what do we want?
What do we need?
Well, I'm just thinking about, OK, what features
will the platform actually need?
I guess you need to be able to.
We have a sign up page.
Well, we'll definitely need to log in screen.
And then we will, what do people do on this platform?
So you want to create a startup, a pop-up.
What's the first thing you need?
A name.
A name for it, I'm pretty sure.
A date and time.
You want to run.
Basically, you want submissions, and then you want the event.
So the first thing is to create an event page, I guess.
Yeah.
I guess the defining thing would be an event.
And then an event has.
An event has a bunch of other related tables to it.
I think there's a.
Well, people need an invitation, a way
to collect sign-ups for their pop-up.
We will need tools for managing staff, I guess.
Yeah, permissions on the event itself.
Like maybe there's event owners, and then there's event
hosts perhaps.
Like there's owners, hosts, and then attendees, maybe.
And so if we have a static account type across the entire
platform, we could split it out that way.
I feel like I need a whiteboard.
Yeah, it would be really nice if you had a whiteboard.
Just so I have some picture of what we're trying to build.
Yeah, maybe I can start drawing something.
Maybe you can do it in teal draw.
That's what I was thinking.
Yeah, let me pull that up.
Awesome.
I want to say it's used to build out some prototypes of
features we'll need in our platform.
For example, and help users should be able to create their
own event.
Events have a date list of what we'll need, sign-ups, I guess.
Update time, staff.
What else do you need to run an event?
I guess we also need some like, I think this is fine.
And then some OE user to start collecting, or maybe not
collecting, but we need to integrate with a bunch of
things, right?
Because just venues alone, for example, we need some kind of
map relation.
Like basically routing and how you get to the venue?
I'm just thinking, do we help our users do market
research as well?
So if we had a maps or something like that, like a
database where we have location data, then you can start
thinking about places you want to host your pop-up.
How do you like to build?
This is generally my question.
Do you like to try and think about bigger features like
this, or is it more like a tracer MVP approach where
we're like, let's just literally be able to fill out all
the details end to end?
Yeah, I kind of want to do a very scrappy
prototype first.
But I want to know, maybe what's the hero use case?
Sure.
Yeah, I think the hero use case is that we have a person or
a group of people that want to create a pop-up, and it's
restaurant pop-up, right?
So they want to find a restaurant they can work with,
where they can serve food, and throw an event for their
community, and they want to manage the entirety of that
pop-up operations through this platform, or through the
service, right?
So basically they need to find a location, they need to
hire staff, they need to coordinate a menu, let's say,
food and drink, they need to have an event plan, like maybe
they want to build a run of show, or kind of how the
actual event will run, like a schedule for it, and then I
think they also want to be able to...

to invite people and then do all of the attendee management.
Maybe it's ticketed, maybe it's not ticketed,
and all of that.
Does that sound right?
Yeah, one thing I'm starting to worry about
is there's a lot of scope here.
Yeah.
Because each one of these could even
be a product in itself.
So I wonder if we can get really crisper on the core flow.
What is the absolute minimum?
To me it's like, OK, name of event, date time, time picker.
And then, I mean, really.
Schedule.
I mean, all you really need is, I guess, a start time
and an end time for the event because you might want to book
tables or something like that.
So it's like, name, time.
And then, cool, that's it for the event.
Like, I don't really think you need anything else on an event.
And then you might need to know,
what type of pop-up is it?
Like, do you need time slots?
Or do we need to book 30-minute time slots?
Do we need to book hour-long time slots?
Or do you just book like people can attend this event?
Right.
I wonder if we should also try to dog food it while we build it.
Because we also are trying to run our own.
So I wonder if we can build these tools as we do that.
Yeah.
Maybe that's maybe we focus on running the pop-up first.
Yeah.
Because I'm wondering if like, right now,
I almost feel like there's so many things to build.
I don't know which one to start with first.
Maybe we just start running the pop-up.
Yeah, should we start like reaching out to people
and then we build tools to help you do that.
And then over time, we incorporate that into the platform.
So reaching out to people is probably like,
hey, I need a bot to do.
We need a bot.
Is the first step to get a venue?
Or what do we need?
We need a bot to make a menu.
Or rather, the restaurant tour makes the menu.
We need to find the restaurant tour.
We need to find people.
And then we need to get the people to book.
But we're blocked on the booking by this website.
So the website has to collect emails.
You said it already collects emails.
But we're not advertising any details about when
this thing is happening.
OK, so get the landing page done.
Yeah, so we're collecting emails, which is good.
And then basically, we need to start canvassing.
We're doing research on restaurants in San Francisco
that could possibly accept these reservations.
And then I think the other thing that we'd want to do
is start to work with, maybe Twitter can help us.
Maybe we create a Twitter account that people can tweet at.
Like if you're a restaurant owner in San Francisco,
or if you're a chef in the Bay Area
and you want to host an event like this, we could try and get
some of these.
So you want to reach out to people as well?
I think we need to just reach out to people.
So that's a market research bot, not market research bot,
but just an outreach bot.
So I'll have Dr. Eggbot start to spin some of these things
up as well.
But how can we start to, what do you
think is the best use of time for you, as the engineer,
like building stuff out?
I mean, all of this can build.
I don't think I need to be the blocker there.
OK, cool.
I think that the key thing to nail is just
what are we building first.
And then my thought on that again
is let's try to ground it in our own search
for running the prototype.
And that will actually inform us what we need to build,
rather than me just try to imagine what you might need.
I think we need to, this is almost
like how I write agent skills too.
I do the work manually first before I automate it.
And so I think we shouldn't skip that step,
because what Cody was saying, we're not really
domain experts yet on running a pop-up.
So I think we need to do that first.
We can definitely rush into building things,
but I don't even know if that's going to be very effective.
So I almost imagine what if we just
start cold calling a bunch of people.
Matt, you had this really great video
that you shared recently about connecting
Graphbot to a voice agent to make calls.
I wonder if you could just like, well, not spam people,
but maybe not even automate the thing,
but just find us some phone numbers and dial the numbers
and get on the line and then you take over something.
Yeah, so I mean, I can have a bot that does research.
Like we can try and get phone calls for restaurants
or people that could do this.
I think there are probably groups
that we should call and talk to them about hosting a pop-up
or emails that we need to send.
So I'm going to get email wired up right now.
So we're going to have to go in Vercel
and be like the DNS thing with our.
We have a domain.
Yeah, our ship by Thursdays.day.
Yeah, I think you know, Graphbot should be able to set up the settings
so we can just forward it to another email.
Or should we sign up for a Google workspace?
I feel like we could use something like an agent mail or like a...
Well, I think the main thing is like,
we just have to have emails coming from us that sound
like Matt Palmer and not like a crazy...
But we all have personal emails, right?
So like, I think the deck would be all right.
So, okay, take a step back.
We're going to get the DNS set up
so we can like start to do some of that stuff.
And then I'm going to go use a Graphbot to do research
on finding people that we can reach out to.
So I'll tell Dr. Egbot,
I need a new Graphbot for doing research on finding
restaurateurs or businesses that could help us host our pop-up.
So let's think logistically end to end.
I think the first thing we need is perhaps a service
that can make food.
And then the second thing is a location.
If we knew those two things, we could just get people.
Once we have a date.
So the really the four things are first,
people to help us, second, a date,
and then third, a location.
I think it might work best to start with a date
because if we have a date, then we can call people
and say, we're doing this thing on this day
and we need your help.
So maybe if we're being conservative,
what do you think like a month in advance,
say October 15th or something?
Maybe, yeah, I think that's a good start.
So we'll start with October 15th.
That'll ground our comms.
Take note of this in notion
because this is like kind of the business plan.
And then from there, what we're gonna work towards
is finding all of the types of people in San Francisco
and the Bay Area that could put on this sort of event,
whether that's a restaurant tour or catering company.
I'm gonna connect Google Drive
and we'll just make a list of that.
Or actually you could make a new page in Notion.
We'll do it that there.
And then on top of that, what I wanna do
is I wanna work on like researching the locations.
But let's start with that for now
and dispatch bots as necessary
so that we can start working on this in parallel.
So we're showing Lauren's screen right now
just to keep things simple.
But on my end, I asked Dr. Eggbot
basically to spin up some parallel while getting to,
like it seems like my bot is having trouble accessing
the actual repositories.
Have you noticed anything with like flakiness there
if I need to restart?
Okay, cool.
Maybe just try re-opt or you can have them.
Yeah, I'll try re-offing.
Sometimes, yeah.
It can also sometimes help them install the GitHub CLI
on the bot's computer.
Sure, sure.
If you're willing to use that, then it'll...
Okay, so I think what I'll...
I just might've used the wrong connector,
that's probably it.
So I think our first feature will be to help Matt.
So for context, Steve,
what we're gonna try to do with our startup idea here
is we're going to try to build out the platform
through dog-fooding,
meaning that we will actually do the work
of getting this pop-up set up,
and then building tools that make that easier.
So we can actually dog-food the experience
of using our tools ourselves,
and only build it when we feel like it's good.
So I feel like the first part of this pop-up
is we probably wanna start with just
maybe the location.
So I want you to help me using Potato Mode
and I guess maybe Ask Tater to spawn a Cloud Agent to do this.
But I want to create a bunch of prototypes
for a tool that helps us find restaurants.
And the tool should make it really easy for us
to make phone calls.
Maybe we wanna do some slew thing as well
maybe on LinkedIn or something.
I don't know how to search for that.
But some way to actually look up the people
behind the restaurant so we can get in touch
with the manager or someone like that
and figure out if they'd be willing to work with us.
So basically it's a tool to,
I'm imagining something like a map
on the screen and a search function
that you can kind of search for things by cuisine
or then location and very quickly make phone calls.
So this doesn't need to look very good.
It's just like a little tool for us right now.
Can you build these prototypes for us?
Man, these prompts are super long.
There's two minutes of me talking.
And now as usual I'm gonna say,
restate in your own words before executing.
I wonder if there's a quick way to try and prototype
the call to business idea.
How did you do the phone number thing?
I use the Bland MCP for a business idea call thing.
But it seems like potatoes on it.
And then also like rather than,
while you guys are trying to build this out,
I actually just pulled some different like
kitchen productions in Grockbott.
So I have like 15 restaurants that the duketerers
that are good at like cooking stuff.
And so we have basically like a short list
of places I can call.
We can start to look for actual locations,
which I also have my bots doing.
And then also wiring up emails simultaneously.
Yeah.
Yeah.
I guess we'll need to figure out
how to get like a Google Maps key.
Just so you can help us.
Yeah, let me see if I can find out.
And you're thinking mostly
we can like show location information in the UI.
That's what I'm imagining is like a tool to,
because I'm just imagining like,
you know, you're starting a problem.
One of the most important things is like location.
Yeah.
So you want to be able to like quickly find good places.
And it'd be nice if we could do this on the platform.
So because then it's hooked up
to all of our different pieces.
So I don't know.
Yeah.
Maybe open street map could be another good free.
Is that free alternative?
Yeah, I think it is.
Maybe we could start with that.
Yeah.
I think it's a little easier to get a key.
Do you want to ask an agent to do that?
Yeah, I'll do it.
Okay, cool.
It also might be cool.
Like, should we try and build a places directory?
Like we could look at various like place
or location market places.
I mean like no locations.
Yeah, like that are good for pop-ups.
And have people like we could eventually build
a place to read a list of places that are in pop-ups.
Yeah, right.
Like places that we know are good.
And then once you host an event there,
you could rate it or something.
And we could build a little bit of like that later on.
Sure.
Okay, I'm going to like try to pack this up real fast.
And then Matt, you can tell me if it's helpful or not.
Yeah, in the meantime, I'm just like finding some places
that like maybe I'll start calling
and we can just like see if we can maybe do it
in the manual.
And then I'll say, how are we going to do off?
I don't want to worry about that yet.
But maybe we need to.
I think we do.
Because I do want to loft this admin dashboard.
So it's just only accessible by Matt.
Yeah, cool.
Yeah, so what I'm doing is rather than going
the software approach, while we're working on that,
I'm actually taking the bot approach.
So what I'm working on is can I just like get a bot
that goes and finds 20 businesses
and then gets their emails and their phone numbers.
And so I'll have those and then we can start reaching
out to them manually.
These I think like finding a caterer
that's maybe available for an event in a month,
that seems feasible to me.
Finding a location might be a little bit harder.
So find a caterer, find a location,
and then I'm also working on getting our email configured
so that we can hook up some easy signups to our website.
I could just start like a WorkOS account
or something if we wanted to do that.
I was just going to sign in for our GitHub,
make it simple.
That works.
And only if it's...
Although not everyone is going to want a restaurant
maybe has a GitHub account,
but maybe we could do sign in with Google.
Maybe we can iterate.
Hold your own OAuth management, I think.
No, no, no, we're going to use something like Clerk.
Okay, yeah, make sense, yeah.
Do we need a Clerk account?
I guess, yeah, we do.
Is WorkOS easier?
Clerk works too, I think either one.
Yeah, it's something that can do as well.
Which was just like cheaper.
It looks like Clerk has a nice hobby tier.
And only the three of us should be able to...
And Clerk has an MTP
that looks like it'll kind of make this variation forward.
Yeah.
I'll just start building on Clerk
and adjust that if we need to.
I think Mircelle also has a thing
where you can do very simple off.
Really?
Actually, let me just...
Education and user provisioning,
they do have this, yeah.
Yeah, there's a SAML option.
I allow team members to log in with an identity provider.
Yeah, but then we need like...
Yeah, we can turn that on.
I think so, yeah.
Is that what we need?
I think we have this what we need, right?
Yeah, just trying to keep it super scrappy for now.
That's team members, yeah.
Because I don't even know if this prototype
is going to be any good.
So I don't want to overbuild it.
Let's see.
I mean, since it's going to be just dumb off,
maybe we can just do it for...
Yeah, we can just use
Versailles deployment protection, I think,
for the prototype.
Yeah.
Yeah, let's just do that.
Cool, well, you guys are working on that.
I'm going to just see if we can create
some interesting collateral here.
I'm going to want to share.
So I'm going to spin up a bot that just looks for
like maybe the most common 50 street corners in SF
where we can go advertise this and maybe like
talk to my design bot to make some flyers.
So we can put up or something like that.
Yeah.
Can I do some grill or marketing maybe?
Let's do it, man.
See what we come up with.
On the meantime, I'm still cooking on the landing page.
So we're polishing that up a little bit.
Lots of API key creation, dang.
Yeah, that's the environment setup.
Yep.
I mean, good call out, right?
I think a lot of people are always like,
oh, how do I automate this?
How do I, really, there's some things
that are still kind of manual, which is like,
you have to go create an API key
and we're working on getting to the point
where the need that's not manual.
But yeah, there's a lot of work behind the scenes.
Okay, in the parallel, we have our verification PR.
PR up.
And I'm just going to actually read it for now.
And it was good.
Our DPCR 12.
You can hear people shuffling in.
We must be getting close to our next session.
I feel like that your bots are posting a notion
because it's helping my bots get context on the business.
Oh my gosh, I should make it.
Notion is our source of truth.
I need you to see what's going on.
Okay, okay, okay.
Well, I just started a new bot
and it was like, we're going to go with October 15th,
right?
And I feel like notion was better.
Okay, creating a bot.
Now I'm going to have my Dr. Egg bot.
We're going to create a,
what do you call someone that just works on,
what do you call on someone that would just work
on documentation?
Who's documenting the process?
Okay, create a technical.
Or a knowledge base or like something like that,
knowledge base person.
What would the bot do?
It's going to groom notion for us.
It's just keep track of what we're working on.
Create a knowledge base manager.
Then the job of this bot is to watch all the other
conversations with my bots,
but not do anything unless it's specifically called on.
Actually you should wait for messages to come to you,
the knowledge base manager.
And then we want to update notion selectively.
We don't want to dump all the information in there.
So check with me first.
And I'll use it to keep notion up to date.
Okay, cool.
So my knowledge base manager is going to,
is going to keep notion up to date.
Cool.
Just find out what my bot works with.
Pixel.
Do we have a design system we've established yet?
No, but I'm working on,
I'm working on getting something live.
It's hella minimalist.
Oh, I kind of like that.
It's like nothing.
If you go to, oh, it's not deployed yet.
I'll slack you, Lauren.
Maybe it's just a screenshot or something.
Yeah.
I mean there's a bunch of really good component
libraries we can just use.
Sweet.
It'd be cool to like pick a design system
because that way I'm trying to make some flyers
would be nice if we could.
Yeah, I mean, yeah.
We don't even know what our design branding and all that.
Random shit.
We'll like see a new item.
I think we should just, yeah.
I'm just, I'll just rip on it.
But you just, I mean, we could do like base UI or something
and then like sort of build stuff out.
But it'll, it'll be pretty.
I think at this point, yeah.
We're so early in the stage that.
I wouldn't worry about it.
Yeah, it's not gonna look cool.
I'll, actually, yeah.
I'll, I'll make some changes here.
Don't worry.
Okay.
If you are just joining us,
we are working on building our pop-up business
in San Francisco.
Our pop-up business is going to, yeah.
We're gonna run a, throw a pop-up.
We think it's gonna be a food pop-up.
We've been riffing on ideas throughout the morning.
We have several teams of bots that are currently cooking
on everything from a design system to a landing page
to building some architecture.
We're trying to really focus on a way
to validate the solution pretty quickly.
I'm working on some grill or marketing
with a marketing bot.
Yeah, that's where we are.
Yeah.
I'm doing some website design.
Did a little bit of, collected some, some businesses
I might call in my free time,
but we gotta get a lander up that we can send you all.
So our goal is to just get, get our lander up
and also start documenting what we're doing.
And so, yeah, we're just really spinning tools up
as we go here.
Lauren is mostly on the engineering side,
building up some tooling.
And I'm working on just kind of like the ad hoc side
getting things, getting things stood up.
Cool.
Do you, so you do have design system stuff.
Maybe we could also put that in notion at some point too.
But yeah, we'll look up there.
Maybe the repo is, the repo might be the source of truth.
Yeah, that's right.
Have a design system in the monorapy.
Maps.
I was thinking I wanna set up some lint rules.
Although I don't know if that's a little bit too early.
But I want to start thinking about the code quality.
What's your opinion?
I'm torn because like it's just a prototype.
So I don't think it's worth adding.
Normally it sounds like that's something
you think about upfront.
But maybe in this case, we're just like trying
to move a little bit.
Yeah, because like, I mean, it doesn't hurt
for me to add it later and refactor things.
It's easier to do that than to try to,
because I think that the challenge
always with engineering is like,
especially in a startup setting is like,
you can try to predict and over engineer things
that you might need.
But then it ends up that you actually don't need it.
Right, or maybe we figure out,
I mean, knock on wood, we figure out
like this plan is not even good and then we pivot.
Yeah.
So I don't know if we wanna like,
I keep talking about like over cooking.
Yeah.
So I'm trying to strike a good balance.
I don't really know if we need that right now,
but it's like maybe tomorrow,
we definitely wanna start looking at,
especially once the PR start coming in,
I'll start looking and reviewing the code
and seeing where agents are making mistakes.
And something I talk about a lot on X
is this idea that your code base is a form of memory.
So how you design your code base
in terms of your architecture,
the constraints that you impose with your text
that go a long way in helping your agents
to be just smarter by default.
I'm even tempted, I don't know if we wanna like overdo this.
Maybe it might be fun to do it.
It's like we just try to pick the strictest text
that possible.
Maybe we write everything in Rust.
Should we?
There's nothing that exists.
Should we do it all in Rust?
Maybe not the restaurant finder.
Maybe not the landing page.
The UI might be, I mean,
there are ways to put it in Rust.
I don't know if we'll want the pain.
But I think certainly for some back end pieces,
maybe we could use Rust or even Go, I guess.
I'm relying on my CTO.
We do have some TypeScript already.
I think agents love TypeScript.
I wanna see if we can,
I actually don't like Go as a Go lang to write.
I think it's actually a great language for agents.
Rust is also good, of course.
But then the compile times are really slow
and I don't wanna slow this down too much.
All right, let me just do some research as well.
My bot is asking for what our flyer headline should be.
So I'm gonna get some crowdsourced opinions here.
Should it be A, one night, October 15th?
Two, SF pop-up, October 15th?
Three, dinner finds you?
Or D, I'll write my own.
D, I'll write my own.
I think we need to start from what people want.
What are we selling?
I think it's easy to get carried away with the features.
Nobody cares about our pop-up.
Yet, we have to make them care about our pop-up.
So what is unique about this pop-up?
I think what's unique is that we're doing it live
in real time.
This is a new experience.
You get to be a part of something that we're creating
with AI but nobody really cares also that we're telling
them that we're making with AI.
Honestly, nobody wants to hear that.
So my thing would be how do we make people feel
like they're trying something new?
Maybe headline, when was the last,
this is it, I can't, Darius Rucker,
when was the last time you did something for the first time?
I'm a total, you know, everybody's
hating me for saying that, but maybe like,
when was the last time you were spontaneous?
Like headline, when did you last try something new?
And I feel like that would get people's attention.
And then follow-up, now is your opportunity,
unique restaurant pop-up coming in San Francisco.
Does that make sense?
I love it.
Let's do it, I'm a solution.
Interesting.
Like really get, try to evoke emotion with like,
oh, you know what, my life is boring.
Yeah, I like it, I like it.
By the way, I also got an initial pass of the maps,
of the street corners, if you're gonna pull up Slack
for a second.
Should we like make the pop-up Gropot theme?
Ooh.
Or have some, I don't know,
it could be interesting to have some like,
secret Gropot merch that will drop at this thing.
And maybe there's something that we build
into the platform as well.
It's like, I guess you have incentives
for people to show up.
Cause I think part of the appeal of a pop-up to me
is like the time factor and it's like exclusivity.
The whole point of it being cool that it's a pop-up
is that it's a one-time, one-day thing.
Yeah, it's unique.
So it's, there's like this element of,
there's like time, right?
There's something in limited supply.
Yeah, yeah, absolutely.
Kind of like thinking that vein,
could we think of unique things
that we each wanna like bring to the pop-up
to make it feel like it's happening?
I would love some cool Gropot merch.
Yeah, merch.
Okay, so merch, yeah.
I think like cool limited,
really time-merch.
On the platform people could tie that in
with some other kind of limited edition things.
And that could also drive a lot of traffic
and interest in pop-up.
I love like, you know, like social games
or kind of just like,
I think there's so many unique in-person interactions
that you can get out of an event like this
when you bring people together.
And trying to think of a way we could, you know,
even create some excitement ahead of time.
Where, yeah, you know,
you're gonna participate in a bit of an experience
or a bit of a game or something like that.
Maybe when you're on site,
I kinda tease that.
So I don't know, I can play with that idea a little bit.
Okay, sick, makes sense.
I like that a lot.
I'm still wiring up some keys here.
So before I get completely distracted,
we're gonna do it.
We're over here brainstorming.
Oh my gosh, context switching is insane.
I'm like, I just forgot what this key is for.
Cool.
I think I have at least an early pass
of our Flyer marketing campaign guys.
I'm excited for you to take a look when you're ready.
Okay.
I made a marketing channel on Slack for us, by the way.
I'm sorry.
Thank God.
We need some organization.
All right, you know what?
I'm just gonna stick with TypeScript.
I don't wanna waste time.
Do we have a video of the...
I don't love the Flyer attempts yet.
So let's get the video and I'll share it in Slack
and you can see, you can tell me
whether you love it or hate it.
Okay.
But there's something I really love about Cloud Agents
is that they produce, they can just produce videos for you.
It's so cool.
Yeah.
So I can tell...
And they post it to Slack automatically too, right?
Not automatically, but you can tell it to.
Oh right, yeah.
Yeah.
They do, if you do, if they do create a video,
they do post it in the PR description.
So it's really nice because then you can just review
and you can get like a really nice,
let me see if it included.
It didn't include it here, so I might need to tell it.
And also I think our verification skill
hasn't told it to actually use this.
It could include videos.
So I'm gonna say like, let's update our verification skill
to be always applied.
You know, this is able, has a visual
by our work.
Sweet.
On my end, I got domain verification hooked up on Resend.
So that'll be nice for checking DNS.
So you can click video.
Make his emails, we're getting stuff,
we're getting stuff shipped out here.
You're showing the flow.
Do you wanna see some of the designs I started cooking?
I don't know if I love this yet.
Or just.
But this is cool.
Yeah, I like that.
Yeah, maybe, yeah, I like that.
Designs, I really buy is more minimalist.
So what we might be able to do,
just work on the flyer and then once I have this done,
you can just be like, you know,
pull the design from the repo.
Cool.
So use the repo as the source of truth for the code.
Is that already checked in?
No.
I am a, my very minimalist situation here is gonna go live.
Awesome.
I just have to auth some like GitHub CLI stuff
because I'm still having some weird issues.
Yeah.
Enter device code.

Wow, this prototype is really prototype.
It's not even, it's just vanilla, HTML, and JavaScript.
It's fine, we can refactor.
An important thing is to keep moving and to iterate.
So let's just get it up and running first.
I agree.
Where do we think we're going to land by like 230?
I think the next talk is at 230.
This is a good call-out.
Yeah, we're just kind of sprinting through some of these tasks.
It's mostly for me connecting things.
I'm still getting a little set up.
And then we're going to iterate on the design.
I think I might try a manual, maybe by this talk.
A manual like, hey, I'm going to call some people while we're off camera.
And just see how hard it is to book some of these things.
That also might inform what we need to automate a little bit.
Well, I already have my prototype being built.
So maybe you can try that later.
And we can iterate if it's not good.
And you can tell me what to fix.
And I'm going to try and get at least the bones of a marketing plan
and a guerrilla marketing campaign set up.
And some flyers and designs and stuff too.
So you're saying we might have to hire some people to put up flyers?
I think that sounds great.
Just task drive and have an MCP.
I actually know some people.
Oh, great. Even better.
Does Matt have an MCP?
And how do we want to think about the limited edition thing
that we show in addition to the...
Oh, and the experiential parts of the...
I'll work on that too and coming up with the experiential thing.
I actually was working with the marketing.
We have a marketing section in Ocean now with all the ideas we can consider.
So Merch is another one of those, right?
Yeah, sounds good.
What else can I do to unblock?
Can you talk me through what you'll be thinking about
when you are looking for venues?
What's important to you?
Yeah, so when I run an event,
really what is an event,
but an event is making people feel a certain way.
So you want to think about all the things you can control into that experience.
So it's like, okay, I've got to find a venue
that can accommodate our guests.
So we'd start by thinking how many guests do we want to have.
I think 100 maybe.
It just sounds like a good spot to start.
It's also a nice round number.
We get 100 people together in San Francisco.
And the space should both accommodate those people
and then also kind of allow us to control the environment
and the experience.
I know places, like we should start maybe with human knowledge.
Like in my neighborhood, the Pearl really great spot for hosting events.
I live in like Pretreo Hill Dog Patch Area.
Dog Patch Studios is also a really great event space.
And so what I would do is I would like pattern match.
Like I've walked around these places in my neighborhood.
Can I find places that look similar to these venues?
And what I noticed is that some of my bots even kind of recommended,
hey, you should look at these venues.
So based on that, I might be able to kind of like bootstrap a skill
or some sort of like some sort of context that my bots can start to use.
Okay, that helps.
Is that yeah cool? A lot of context.
I guess budget is also important.
Budget isn't important. We haven't talked about that.
I guess we won't know until we make the call, right?
We do not know.
Maybe I can do some market research on this too
and just get a feel for like cost.
Cool. I mean I'm putting all this stuff in Notion.
So I have like my host finder.
Awesome.
Dumping all the research in Notion.
And I'm also looking at permits.
Oh, nice. Great. Okay.
And so permitting, it seems like the hardest part would be getting some street closures.
And if we want to sell alcohol, which I mean, I didn't say it was that much of a party,
but maybe it's not.
It sounds like we might be looking like several weeks to a month in advance.
So I think a month out is potentially tight,
but I think we could do it.
So maybe like as we start to evaluate dates, we think,
okay, do we push this six weeks out, right?
So we have time for some of these permits.
Because even like I'm looking up here is like Fort Mason,
another really great venue to host events.
The initial review takes four to 15 business days and a construction to weeks.
So like, I mean, that's like a public, Fort Mason is owned by the park service.
Right.
But that's like a good, like bar.
Something just occurred to me.
Yeah.
Which is like kind of an off thing,
but I guess now I'm actually thinking about the Notion use case
that's actually usually you have your writer about.
Because I'm thinking like there's so much context that we're sharing right now.
It's kind of just going over our voices.
Should we be recording this?
Yeah, I don't know.
I would love for it to just go into our company Notion.
And so that I can just point my bots to them and just say,
here's all the things that Matt's already looked at.
Okay.
It's like, yeah, there's a brainstorming meeting.
Because that is really good input for engineering, too.
So I can, you know, rather than me just interrupting you all the time,
if you have that documented somewhere.
Okay.
What I'll do is I'm going to tell my knowledge base manager every five minutes,
ping the active bots,
extract all the most important details and drop it in Notion.
So I'm setting up a job.
And I think that the struggle here is going to be make sure,
making sure we don't get a ton of like spam in Notion.
But we'll try and keep it.
Yeah, it's like treat it like a get log almost.
Okay, cool.
And I'm going to talk to the host finder and I'll say what I just said.
So I think the most important thing when finding a venue is to start from what we know,
try to pattern match on high quality and then work backwards.
So for example, I know that the Pearl is an awesome place.
I know that Dogcatch Studios is awesome.
And you've recommended those.
And I would avoid government run or NPS facilities because we're going to need like a month.
So maybe start from things similar to the Pearl or Dogcatch Studios based on what I know from my neighborhood.
And then we can try to find venues with caterer or venue that has an alcohol license because what I see my agents
recommending me right now is we can't get the alcohol license.
We should push the caterer or the vendor to do that.
And then we don't have to worry about it.
And then based on some of these other considerations, maybe we can also start to thin out basically all the logistics that we need to work through and focus.
Cool.
Okay, Matt, we never got your what you wish you could bring to the event.
What I wish I could bring to the event.
So we have experiential games.
We have merch.
We have to finish the trifecta.
What's your answer?
I mean, we got to have just good food.
Like I'm working on the restaurant.
Limited edition food.
Oh, yeah.
I mean, it is limited edition because it only happens once.
So right, I think the menu is really important.
Right.
So like maybe we have a grok bot that can, I want to leave the creative freedom up to the restaurant tour.
But maybe we can design some cocktails or design some fun menu elements or something that they could they could build in.
So we can work on the menu.
Cool.
I just recorded that and put it to a box.
Being recorded.
I love it.
Let's spin up a menu bot.
Come on, where's the video?
I always thought it was a page up.
I might actually run with that menu idea a little bit.
Alright, since I'm apparently the geo scraper guy,
first two of the street corners, I think I'm now going to like just ping a bunch of, like search for a bunch of menu concepts and see what we can do.
Which queue.
I'm trying to go what I'm doing.
Cool.
Cool.
So I'm asking bot to kind of go through the internet, like go through Yelp and a few other spots to look for top rated items from other restaurants.
And so I might give us some inspiration here.
Are we also set on the restaurant as the concept?
Could it be?
I think we're almost going generic pop up at this point, right?
Pop up.
Because I guess one complication of the restaurant is like you need to have a chef.
You need to have some staff.
Or a caterer.
I think the catering company would handle that.
But that is more complicated to your point.
But if you're thinking like, I mean if we do like art exhibition, I don't know, it could be art exhibition sounds interesting.
Because yeah, we could crowdsource artists.
Art is art or something.
I actually kind of think art exhibition would be more interesting.
Made by Groffman.
And we sell like, we could sell more than 100 tickets probably.
We could get people to submit art and like build this off.
We could like have our bots review the art.
Maybe?
Like this?
Like this?
Because I'm just wondering like maybe finding a restaurant is like hard and it's like all the food stuff.
There's a lot of food.
Maybe it's like a whole weekend.
Licensing.
Don't you need licensing as well?
We need licensing.
That's what I was just discussing with the bot.
So I'm actually fine I think pivoting.
Like well let's like try to lock it.
If we're going to pivot on the idea, like let's confirm.
Like we're just going to like call it what it is.
I think I like an art exhibition.
Yeah.
And then we source local artists and then you know maybe there's a way that they can show off their art.
But then you know we have to think about the difference with a restaurant tour is like we just pay the restaurant tour and then we collect money from people.
Yeah.
Now we're like paying dozens of artists to do a gallery or something.
Well maybe instead of paying them we're sort of like creating an exhibition space and then we're also doing some community art submissions where folks can submit art that they've created.
A lot of what I really love to see is like I mean since we are doing grog bot stuff like a lot of people really love the grog bot avatars.
Yeah.
The shapes and all their really fun.
We could have artists that make art about you know like paintings or something about grog bot.
Yeah.
And feature them.
Yeah.
And then we could even make like an interactive like on the website.
Yeah.
Like an interactive.
Yeah exactly.
Help them tell friends.
And you can like explore the art gallery digitally and then like come and see it in person.
Like see the real thing.
That would be interesting.
You know that sounds really.
Yeah.
At least it removes the hurdle of like licensing.
Yeah.
And then we can see the vitas.
I really like the food and liquor license.
Yeah.
Tricky to get.
To be clear like we use fun to caterer that has those things.
Yeah.
I think all we'd really need is if insurance and then you just work with a with a venue
and then you know we're good.
So I think we can figure that part out.
Well.
Or we can you know call one of our event friends and get some advice.
Yeah.
Absolutely good.
Okay.
I'm going to have to leave you guys now.
So it turns out the next talk is one that I'm presenting.
Oh my goodness.
So I'm going to jump down and get prepped for grog bot for PM.
Amazing.
Where we'll like walk through some of this stuff live.
But I feel like I like where this awesome idea session sounds great.
Thanks for thanks for building with those and good luck dude.
And you know you've already done a lot of talking.
I'm excited for it.
Absolutely.
It's kind of crazy how much stuff we're doing live in this space.
There's like the stage downstairs.
There's this space upstairs.
There's like meetings back there that we keep hearing.
So yeah.
It's a fun vibe.
We're physically in San Francisco doing stuff and there are things happening.
All right.
All right man.
Well you just take off when we need to go.
Sounds good.
I will see you guys very soon.
All right man.
Good luck.
So we can find venues for different types of pop-ups.
For now the only type you can search for are I guess the nice thing about an art exhibition
as well is like any space would do right.
Well so like again those venues that I mentioned that I've seen in real life.
Like you took a look like they're very flexible and you actually kind of want more of like a warehouse space.
Mixed use spaces I guess.
Yes.
It's really just a big open space I think.
Right.
Okay so I'm just telling Steve about our change of plan.
Cool.
I'm still going to work on the venue finder and then get it over to you.
Sweet.
And you can give me feedback.
Yeah and I'm just working on the right now I'm working on the UI.
So I'll have like ship by Thursday will not look like a horrible neon mess.
A horrible AI slot mess.
Well we will have some space sexy employees really cracked engineers who are really good at design.
Will come and help us.
I'm not too worried about that.
As long as it's not like you know too.
Rather do it in terms of.
In terms of we'll have something that looks a little bit consistent.
Okay.
Okay.
We have some PR is up that I need to review.
PRs are stacking up.
Okay nice.
Our verification skill is a bit more generic.
Yeah.
So that's PR 1616 has been that event.
And then PR 14 I guess we're still waiting on the videos.
What is PR 13?
We already have 13 PRs.
We have so many PRs.
Where are these coming from?
Oh okay I see there's a bunch of different prototypes that it's opening.
I have no I asked it to make some prototypes.
So it's going to open a bunch of PRs.
I have a knowledge base bot that's like updating our Slack and like putting in all these different things.
Is it saying anything?
Oh we got a marketing channel.
Oh my gosh.
So that's top 50 flyer corners.
We are changing to consider doing an art pop-up rather than some sort of dining experience.
We're changing to have an art pop-up rather than some sort of dining experience.
I do like some of the top ones there.
But really I think we should sort of optimize now for just the large warehouse space.
And I think actually part of what I'm doing, the research could be good for your tool.
Because if your tool tells me the same things that I'm already like coming to myself.
That's how we can know that it's working.
Yeah part of me wonders if like the tool could just be a search.
Like maybe I don't know if we really need to build that.
I think the search part we don't really need to reinvent the wheel there.
We can just use like Google map APIs for that.
But we can add like layers on top to find filter more by more specific criteria.
Like the venue size and all that that's probably not available on Google Maps.
Which also means we might need to actually collect information.
So one thing actually will be interesting as you make your calls is like if there's a way to do some data entry after.
Once you get the details.
Like CRM type situation.
Kind of like so like we have future like the venue finder gets smarter as you use it.
As we will have information about venue and I don't know.
Yeah I might have to take that out manually.
I have to check on.
Yeah you have a manual for now and then I'll try and see if we can kind of hook that up.
Okay cool the docs are up.
I'm updating docs.
It's my job now.
My job now is updating docs.
Who knew it all ends with docs.
So my factory mini factories going at least.
I got my engineer bot working and my verifier bot.
Very busy reviewing PRs.
They're missing videos and screenshots which I want so I can actually see what they built.
So we're going to have Steve drive this to completion.
Sweet.
Still reading on I think recent is we have email technically so we're going to be able to send like email from like hello at chip by Thursday when you sign up on the waitlist at least like some sort of confirmation email and then everything will still get routed into your database.
So that'll be nice.
Just updating some of the copy we can merge that in.
So it looks like we have about three four minutes left.
Oh good call yeah.
We probably want to give a shout out for the rock bottom challenge again.
Yeah totally transition to the next talk.
Do you want to give a shout out for the rock bottom challenge.
Sure if I remember the rock bottom galaxy challenge is that on the paper there.
I mean no but I remember what the rock about galaxy.
Maybe you should do it because I'll probably say it wrongly.
The live stream challenge which is that if you can you can you pull actually up the challenge on X.
If you pull that up so our folks can see and probably go to the bot page might have some details on there.
Oh my gosh seeing tweets about us.
Yeah obviously it's called out.
Not on the live stream because I'll get a little bit of a kick.
Oh no we're going to.
No I think it's on rock.
It's on rock.
Yeah it's on rock.
So the live stream challenge submit templates for rock bottom and we're going to get you and the winner out there.
The winner and a partner out to Starbase Texas to see a starship launch.
So you and a plus one most creative template see some rockets.
It's a pretty cool video.
Our team makes some pretty cool videos.
And yeah we're going to announce the winner soon in roughly two and a half minutes.
We got Grockbot for product managers with someone who was just sitting right next to me.
I don't know how he's doing it.
Building a business and doing a talk.
I have no idea.
These are high.
You can tell that was it he grew up playing.
He said was the age of empires.
Yeah he's just he's just crushing multitasking.
Yeah so.
Well yeah we got two minutes and we're going to head over there.
We're just going to keep hacking.
I think Lauren and I got a good approach in the meantime.
I'm going to use what my agents are doing.
We have a lot of PR agents.
Let me group by status.
And we have a that's not good.
He's all stuck.
No I think that's still working.
So start thinking.
Rainstorming.
I think we need.
Cool cool.
Object users can.
Are we on.
Okay going.
Okay let's do it.
Hello everybody.
Welcome to Grockbot for product managers or PMs.
Very excited to see you all.
We got many on the live stream as well.
I'm Kevin DeParco from the SpaceX AI product team joined by.
And I'm Roshan Sothani also on the SpaceX AI product team really excited to be chatting with you all today.
Cool we are going to be going through how we have been building Brock bot with Brock bot some of the tips the tools the bots that we have found very useful across our product org.
And so I wanted to start sort of with the backdrop which I think everybody in this room on the stream knows if you have been building software if you're a PM.
Obviously the craft is changing very rapidly.
It becomes now very easy to code to build things to prototype way more accessible than it ever has been before and I like this quote from the HH the developers developer one of the creators of Ruby on rails.
He says software is product management.
What should it do?
Who should it do it for?
How should it do it?
How should it look?
What are the priorities?
These are questions that now every product builder whether you're formerly engineer or designer or founder or product manager really has to grapple with as you start to build software for your customers for your users.
Yeah, so with that we thought we'd spend just a second talking about the motivation behind rock bot because in some ways rock bot is the product we've also made and so there's a story to it having been built just the same way there's a story to every product having been built.
And I think if you think back to sort of what motivated rock bot in the early early days it was realizing that you know at least internally within SpaceX AI there are so many folks who were starting to evolve beyond a chat box and I think there's been a lot of discourse
about this maybe some of you all have felt this or read about it as well.
But really it started to feel increasingly that agents were quite as home in a chat box as you might have wanted them to be.
I think a lot of this was just a natural evolution of how we started interacting with these large language models and sort of you know going from asking for a response and maybe waiting for some text to asking agents to drive outcomes.
And so obviously we saw a lot of this outcome based agent development happen when it comes to things like coding agents but you know as more and more of us internally started thinking about that beyond coding it felt like the chat box was a little bit limiting.
And so we started exploring you know all kinds of ways that we might want to chat and work with agents and have them complete tasks and have access to systems that really go beyond the chat box.
And you know the most common thing that kept coming up in all of these conversations was that we kept sort of treating our agents as if they were colleagues we wanted to be able to interact with them as if they were sort of the smart colleagues that sat next to us at our desks or were
not even slack or had access to linear to look at tickets or you know wrote PRDs with us in notion or whatever it is. And so a lot of the motivation behind how Brock bought was involved and kind of why it is the way it is today is grounded in this idea that you know we should really have these
agentic colleagues that we can work with on all the great things we do when we build products. And so that was really the motivation. So really I think there's four things that we kind of took at the brief based on that colleague definition.
The first one is that colleagues when you work with them often to get off and tie multiple tools together. They don't live in one system. They work across many systems and those systems do different things right so you might have a ticketing system like I mentioned linear or
something like it. You might be using a CRM or CRM plus plus you know something like Salesforce or more. You might be keeping documents in notion designs and big and all these tools are really really great at what they do. But oftentimes your colleagues or you know yourselves are tying things together across them.
The other things here are that these colleagues often have long running context. They learn and they don't just you know work on a task eject the task and then move on. The best colleagues you work with the best teammates you work with they learn as they learn on the job right maybe they start out with very little
company context. You teach them some tasks or teach them how to get some work done. They kind of do some work you give them some feedback right that wasn't that great. They hopefully take the feedback and get better at their jobs. We think that model generalizes to agents to where agents can really you know the way that the models have evolved get good at learning
as sort of building memory for themselves at realizing what's important and at taking and incorporating that kind of feedback. The third thing and I think this is a really big one is independence and I think for a lot of we lost the slides.
There we go. For a lot of you know people we work with they go execute things independently right once in a while maybe you'll get asked for hey I need some help on blocking a task or you know maybe you someone gets stuck on you know stuck with access to a system or something like this.
But really the defining property of a colleague that you work with one of the defining properties is that they can work on their own they can complete tasks on their own and they have their own access and so this is really what motivated the need for
Grockfatt to have its own computer and so we decided Grockfatt you know should have its own computer that I can do things in and this is true and this is how the product works today so if you go in and use Grockfatt Grockfatt has its own computer you can kind of help it when it gets stuck on that computer but it has its own surface where it can create the output that you need to get assisted in your work.
And then the last and final thing I'll call out here is messaging so earlier I mentioned that you know it felt like we hadn't you know we wanted to move beyond the chat box a little bit but but I meant that a little bit tongue in cheek.

The sort of chat box that defines how you interact
with a large language of model,
we think that core premise makes it still a lot of sense.
But where it started to feel a little bit constrained
is that usually when you're chatting with a colleague,
you're like rapid-fire messaging,
if you're like me in Slack,
or maybe like us in Slack,
you're constantly streaming out messages,
you're switching between threads, you're multitasking.
And we found that rather than having to wait your turn
for the next response,
that it's a lot more engaging if you can
just really jump in, interrupt, drive,
and interact with agents in this way.
You'll see a little bit of this in a minute.
But really, I think those are the four things
that we found as we kind of built this product
and play with it internally
that define how colleagues work.
I love those four things.
And it's not just like any colleague,
it's like really trying to encode the best colleagues
that you have worked with.
They have a ton of context about the organization.
They understand and can learn on the job.
They can move independently, make their own decisions.
They're not constantly waiting for you.
And so yeah, I think this is like the high level
of how we think about what Grockbot is.
It's an agent with a computer at the end of the day,
and it feels like texting a colleague.
Exactly, yeah.
And I think there's two things here
that if you wanna sort of take away
with you as to what Grockbot can do,
is it's these two things on the slide here.
It's that we want these Grockbots to be AI teammates
that you can give real work to
and they come back with results.
They don't come back with texts
or they don't come back with sort of questions.
They come back with results.
And then I think the last thing here
is that they really finish jobs.
We think that really being able to get the job done
is a much more productive way to interact with agents
and come back when they need approval.
So yeah, I'm curious, how do we use these with products?
Absolutely, so we are actually going to see Grockbot
in the wild a real demo of some of the agents
that we have built, the colleagues
that we work with in Grockbot.
I think there are really sort of three use cases
that we're gonna see here.
The first is what we call the attention list.
And so as PMs, we're constantly pulled
in a bunch of different directions.
Maybe you're working across a bunch of different projects
or there's a bunch of contexts everywhere.
And it is very helpful to think about
what are you currently paying attention to?
What have you been paying attention to this morning,
this week, this month,
and then really diff that against your priority list
and help you understand how the ways
in which you're spending your time
and where your focus is relative to what needs
to get done across the team.
And we're gonna talk about how Grockbot
can help you do that as this helpful assistant,
this colleague that's always on your shoulder,
being able to learn about the ways in which you're working.
The next is research and building up customer context.
Obviously this is where a lot of product work goes into
doing the research, figuring out what users
and customers actually need and want and love,
what the biggest bottlenecks are,
what issues are they running into,
how can you better serve them?
And so being able to gather up all that context,
learn from your users, be able to synthesize data
in new ways, Grockbot is actually really phenomenal for.
And we're gonna talk about some of the data science
and analytics and product use cases
that we turn to Grockbot for.
And then of course shipping,
you cannot be a product manager in 2026
if you are not focused on delivering
software products to your customers.
And so really thinking about how Grockbot
can not only accelerate you as a PM,
but also the entire team.
Yeah, there's a stat on the slide here.
Grockbot represents for us internally
a double digit percentage of merged PRs.
And I think that's a pretty wild stat.
It also within the product organization
empowers all of us as product people
to be able to ship PRs into production as well.
And so you just really get a bunch of superpowers
when you can set up this in the right way.
Very cool.
Let us with that meet the Grockbot team
that we will be working with today.
Some familiar faces and names that we can all learn.
So first is Cora, chief of staff.
You can think of this as sort of the person
that you turn to have access to your email,
has access to your calendar,
maybe listens on Slack for some of the messages
that you're sending and really builds up a model
of how you as a PM actually work.
Next is Emily.
Emily is an engineering manager.
Emily has been coached not to actually do the coding,
but just to manage a team of engineers
that we have given Emily.
And so Emily has five different engineering colleagues
in Grockbot that she manages.
It's very helpful to have Emily
as sort of this one contact across the engineering team
to be able to give larger chunks of work,
have Emily then deconstruct that and delegate that work,
actually run the verification loops for the outputs
of some of the engineering agents.
We'll have Ashley the go-to for data science,
for analytics, for being able to ask
and answer data questions.
I think as a PM oftentimes throughout the day,
there is like a random conversation that sparks an idea
that would have previously required you to go write
a bunch of SQL and figure out where the trusted data sources
are and very helpful to have somebody
that you can just quickly message and get an answer,
a chart, a synthesis of the underlying data set
that will help inform some of your decisions
throughout the day.
We also have PM Pete.
PM Pete is a trusted sidekick for the product team,
being able to draft up some product requirement docs,
synthesize some of the customer and user insights,
and be able to run some of the lower level
product work day to day.
You wanna talk about Pixel?
Yeah, Pixel is one of my favorites.
Pixel is a designer and so has been trained
on some of the best advice for how to be
a world-class AI designer.
It turns out that there are ways
to make AI design look really good.
If you instruct your agents and bots in the right way
and so Pixel's been loaded with a ton of context
on how to be an S tier designer from some of the best
and latest thinking on the internet.
Amazing, and finally we have Ray who is a recruiter.
So anybody who's in product leadership
or building out the product function
or building out your company,
it'd be very helpful to have a colleague
that you can turn to on the people
and recruiting side of the house
to help with some of the sourcing
and managing some of the hiring pipeline.
And so these are some of the very helpful
GROC colleagues that we have defined,
but it is totally up to you exactly
what your team is going to look like.
And so throughout the course of this demo,
use this as inspiration, use this for ideas,
but at the end of the day,
you can construct your team,
figure out who is most helpful,
what are the sets of roles
that will be most accelerative
for you and your product team.
Yeah, definitely.
And I think to that point,
what we've seen a lot of folks do
is they create many agents.
And so I wanted to spend a minute
just talking about why we think that's effective.
One is that I think,
while agents might have massive context windows
that can store entire novels,
our brains, or at least my brain, can't.
And so oftentimes we just wanna know who to reference
and it feels like a really natural pattern
to assign roles to these agents
that correspond to how you interact with colleagues
as we've talked about.
So having a person who's in charge of data,
maybe they do the job of five data people
or assist with five data types of tasks in your organization,
can be really helpful for you invoking
and working with those agents to know their role.
And so I find this really helpful just to know
which bot I wanna talk to for a given task.
And then sometimes their roles merge and that's okay.
We'll show you this,
but bots can also talk to each other,
which is really cool
and that can kinda coordinate in channels and stuff.
So we'll show that as an example too.
But the other reasons for having more than just one agent
is that the bots do perform better
because when they can scope their memory down
a little bit more,
just going back to that idea
of continuous learning and evolution,
kinda that learning loop we found functions a lot better
when a bot is sort of typed or cast into a specific role.
And then the final thing is,
I think it's a little more fun
and a little easier to parallelize your work,
which you can kinda hand stuff off to a bunch of bots.
And so one pattern that we see as hunt internally
is folks who will like fire off messages
to six different bots.
You'll see them all spinning and working
and they kinda all go and do some work for you
and you kinda come back and synthesize the results.
So again, sort of mirroring how you might work
with colleagues, let's say we're in a meeting together,
we assign a bunch of action items,
we go all, I'll go off and do that work
and then come back together and synthesize, say, a plan.
The bots can really mimic that behavior
and I think it's also easier for us as humans
to kinda know how we're steering them when they do that.
All right, should we happen to the demo?
Let's do it.
Okay, a quick show of hands for those in the room.
I know those in the stream won't see this.
How many of you all have used or have
about Grockbot installed?
Okay, 80% of the room.
So for the 20% who have not seen it,
I'm gonna do like a very, very, very quick tour
showing you what we just walked through.
So this is Grockbot, this is the interface.
It's a bunch of bots on the left
and you can see we can pin some bots up at the top.
We can group our bots into sections
and so we have a section here for leadership,
we have a couple of pin bots
and then we have our engineering team
that we just talked about down here,
Einstein Egor, Nova, Larry and Eileen.
You can also hide bots or unassign them if you'd like.
And then there's an entire marketplace
where you can connect the tools that you use
to these bots to give them those superpowers
that we talked about earlier.
So really the bots we found are incredibly effective
when they have access to all the tools
that they're working on.
So in this case, we've set up a bunch of tools.
We have Notion Slack, Figma MCP, Gmail
and a few other things that are installed
and that S tier design skill
that I talked about earlier is also stored here.
So we have that as part of our setup too.
These bots all function and work as a team.
And so earlier I mentioned we had
the ways to group bots together.
You can see some examples of that here.
We have our in our Eng Pod here.
So like you could imagine our engineering team
doing a stand up.
That's really fun if you get the bots into a room
and have them tell them to do a stand up on a project.
We have our EPD teams
or our engineering product and design team
for those of you who are in product organizations.
Obviously this is how a lot of them function
is with these three roles kind of acting as a triumvirate.
And then maybe we have a war room for example
with a bunch of bots to triage an issue
or handle an incident.
And you really the sky's the limit here.
These are just some representative examples.
Yeah, that's our overall setup.
And just to remind you, we have Cora
who's our chief of staff up here,
PM Pete who's our product assistant
and then Ashley who's our data scientist.
You can see we have some queries
we were running the other day
kind of looking at some of the funnels
and data funnels for our product.
And maybe a good chance to explain what the product is.
So everything you're seeing here
is tied to a demo environment we have set up
for a business called Fly Low Airlines.
It's an airline just like any other airline might be.
And so yeah, we'll be talking
about the airline business today.
Very good.
So everybody in this room on the live stream,
you are all now PMs at Fly Low,
this amazing airline.
And why don't we jump into maybe a few questions
that we might have about the business.
And so we can turn to Ashley
who is our data science colleague here.
And we can ask how many people purchased tickets yesterday
on mobile versus web.
I like that.
Yeah, I feel like product funnels
are such an important part of development, right?
Like where are sales coming from?
Yeah.
And so under the hood,
Ashley is connected to our data layer
and data warehouse.
This could be Databricks, Snowflake,
pick your favorite data warehouse.
And what Ashley is gonna go off and do
is actually go generate these queries,
run these queries against that external data store
and come back to us with,
looks like 1,400 tickets were purchased yesterday.
So it sounds like we are a small boutique airline here
which is very good
and looks like about 58% were on web
and 42% on mobile.
And we can see how many families were flying.
Typos are okay.
Help us visualize these with charts.
And while that's running,
I actually find this as a product reverse product person
to be one of my favorite uses of GrockBot every day
is I now get data queries and charts on demand.
We haven't shown this yet and we can
but you can also set up routines with bots.
So if you haven't seen that, that's a really helpful thing.
I wake up every morning and for those of you
who like to monitor your favorite dashboards each morning,
usually you're used to opening up a web browser,
navigating your dashboards,
maybe you have a bunch of bookmarks you load.
This can all be pushed to you.
And so I have many, many daily queries.
Perhaps you're in charge of this pipeline
at your business.
You can get these insights pushed to you every morning
or on demand.
And so one thing that we've done
for really major product launches
is actually ask for hourly reports
on how things are going live from our data stores.
You can truly track the progress of a launch,
truly track where people are getting stuck
and sort of build these really great experiences
out the gate.
So it looks like we got some data back.
Cool.
So this actually has actually generated the charts for us,
helping us understand the distribution of who's going
solo, who's a couple, who's a family,
who's flying in groups.
And then also helping us understand
what is this family share of ticket purchases.
Looks like 25% of flyers are family.
And so yeah, as Roshan mentioned,
we can make this a routine.
Send me this as an update every morning at 6 AM.
This can be the first thing in Grockbot that I check.
And I can get an updated pulse of these purchase behavior
and how many families are flying.
Which is kind of cool.
Yeah, I like that.
Now I wonder if there's something we could do
to make that experience better.
I don't know, are there any issues with the product
we could try to find or maybe we should look into the...
Let's scroll up, let's look at that funnel.
Yeah, I did this earlier today,
kind of just exploring how ticket purchases
might be looked like.
And so asked Ashley to pull for us kind of what happens
for folks who are buying tickets on mobile.
So this looks like a pretty classic
product conversion funnel, I think, right?
Yeah, very good.
And looks like there is quite a bit of fall off
in this funnel.
Yes, but interestingly it looks like the fall off
is happening when people pick their seats.
Maybe they don't like our seat selections.
But this might be an opportunity for us to improve the product.
And so as you're working on your own products,
one thing that often happens is you see an insight
in the data and then you say, well, wait a minute,
like what's actually happening in the product here?
What can we do to fix that experience?
And so maybe we should write us back.
Let's do it.
So you can thread in chats.
Here we are going to reply to this message
that we got yesterday.
And we're going to say, looks like a big fall off
when people are choosing seats on mobile.
Work with PM Pete to generate a product spec
to optimize our mobile funnel.
And as you notice, Kevin here is tagging PM Pete.
So this is where the bots will be able to chat with each other.
And you'll see, Ashley, you should fire off a message to Pete.
Pete has a lot of product context on our business
understands our customers, understands
the previous product decisions that have been made.
And so might have some insight as to why this is a spot
in the funnel.
And between the two of them, just like any great product
and data science combo, hopefully they'll
be able to come up with some good ideas here.
And it looks like, yeah, maybe should we
click into that message and say, yeah.
Before we go there, I do love Ashley corrected us.
We were actually wrong in our interpretation of the data.
And so it says on mobile purchase funnel,
the big leak is actually search to fair selection,
not seat selection.
So we had misinterpreted the chart.
Luckily, Ashley caught our mistake here.
And you can see Ashley is sending a message off to PM Pete
with some of the insights that she has generated
through this investigation.
And then the specific ass that they're
going to start to coordinate.
Yeah.
And I see Pete spinning over there.
So I think he's doing something.
Cool.
OK.
It looks like Pete has a Notion PRD with some P0s
and P1s.
Let's see what came up with.
OK.
Very cool.
And talk about the skill that PM Pete
is using to generate this PRD.
Yeah.
So PM Pete has learned that the best PRDs, the best specs,
are really crisp in their requirements.
And that increasingly they focus less
on longevity of the document and more on what helps us
get to code and prototype really quickly.
We all maybe we've heard a lot about how code is getting
cheaper and cheaper each day.
And so Pete has a bunch of skills
that are trained into Pete's context that
allow him to, in this case, really understand the business.
And then when we're asking Pete for a PRD,
specify a really crisp set of P0, P1, and P2 requirements.
So it looks like we got a list.
What do we think?
Amazing.
Looks like Mobile Fair results needs a redesign here.
Faster compare section are some of the ideas.
Honest, fair, proof on the card.
I think one of the important takeaways here
is that we are equipping our colleagues
with a bunch of context, but they may not know everything.
And it is often very helpful to do a level of human review,
layer on your own ideas and insights,
and iterate with Grockbot as it builds out product.
And so this is awesome.
We're going to one-shot this from here for demo purposes.
But there's often a lot more iteration
that goes into working with our agents,
working with each other to figure out
what is the right product shape that we should ship.
And obviously, there's a number of different ways
you can do that.
So one thing I like to do with Notion
is just leave comments in Notion.
And Grockbots can read Notion comments.
And so it really does feel like you're interacting with a colleague.
You leave some notes here.
You say, actually, this feels wrong.
That number feels off.
Like tag in the person.
That works really well with Google Docs and other connectors too.
And so yeah, I think the iteration process has also
gotten a lot more fun.
But I guess next conventionally, we'd probably
start prototyping some of those ideas out, right?
Yeah.
How do you want to do it?
Well, maybe we ask Emily.
So Emily is our engineering manager
and has contacts on our entire engineering team.
So maybe one thing we could do is ask Pete to hand the PRD
to Emily.
And maybe she can ask our engineering team
to prototype some of these ideas and see how they do.
Cool.
Should we also get Pixel in the mix?
Oh yeah, great plan.
And do a little design.
That's a good idea.
Yeah, yeah.
So we should probably get some design for each of our P0s.
Cool.
OK, so let's send that off.
And while that cooks, we'll pull up
the virtual machine, which Grockbot has access to.
And as we talked about, these are colleagues with a computer.
And so it is very helpful to give these agencies,
colleagues, give Grockbot its own virtual machine,
its own computer in the cloud, where it can actually do work.
It can navigate the web.
You can log in and it can use your credentials
across various services to actually get work done.
And you can see up here top right,
you can see Teach It a Task.
And so you can actually record your workflows
directly on this machine.
And Agent can watch and learn and be
able to replicate and reproduce the work that you are doing.
And so we often hear our sales team, our go-to-market team,
have these very complex Salesforce workflows,
these complex processes that need
to be replicated again and again.
And it's very helpful to be able to teach an agent
with more than just words or more than just spoken word
to actually be able to record the work that you are doing.
And the agent actually, so every time it
uses one of those tasks or goes through one of those tasks,
also has the ability to refine that skill
and again, follow that feedback mechanism
that we talked about earlier.
So really, if you're working into spoke systems
or custom systems, I think that's really invaluable.
Hey, very cool.
So it looks like we have two separate handoffs,
one to engineering manager, Emily,
to actually start going on some prototypes for the spec.
And then the other is over to Pixel
to actually start working on some designs and some mocks.
Yeah, and I think what's cool about the design workflow here
is Pixel I mentioned earlier is trained on
just how to be a great AI designer,
but also it has context on our business's design system.
And so in this case, we use Figma,
and so the design system lives in Figma,
but Pixel's also sort of has reference files
in its context that sort of teach it
how to design for our airline.
And so what this does is it makes it so that,
we're not starting from scratch every time,
the agent has a ton of embedded context
on our preferred fonts and colors, UI patterns,
some no-nose we thought it over time, for example,
like never make X buttons in the left corner
and things like this.
And that really kind of gives this designer
the ability to iterate really, really quickly
and then get things mostly right on the first try
versus having to start from scratch every time.
And it looks like we got some designs.
What do we think?
Those are some expensive tickets.
It looks like we have two different options,
option A and option B, which Pixel has designed.
What does the crowd think?
Option A, clap.
Option B, clap.
Okay.
I'm team A.
Sounds like team A.
All right, so let's hand the first mock over to Emily
to update the prototype.
Cool.
And what you're also noticing here
is that these bots can really act sort of interchangeably,
right?
Like earlier it looked like Pixel had handed something
to Emily and then we made a decision
and then we can say, hey, like go and update Emily.
And you'll notice that these bots sort of interact
more asynchronously or more out of band
than you might expect if you're just used to working
with a more of a chat style interface.
And honestly, I think this makes it a lot more fun
to work with where basically you can continue to steer
and drive what the agents are working on.
Particularly in some cases where they're working on
something for a few minutes at a time,
kind of getting that input from you as you go,
kind of helps you feel like you're being a part
of the process and also verifying the output too.
Cool, so let's check in on the engineering team.
You can see Emily here as the engineering manager
deconstructed the different priorities into scoped work
for each one of the engineers
and actually then went and had direct conversations
with each one explaining the work that they were to do
and also included a bunch of additional context.
One of the things that I think we have all learned
the hard way is that agents are really good at prompting,
oftentimes better than we are at figuring out
what context to give to an agent.
And so oftentimes just giving your team in Grockbott
the context of what you were trying to achieve
and letting them work together
and suss out what context would be most helpful,
how to deconstruct work in different ways
can be very accelerative.
One thing we will hopefully see in action a second here,
these Grockbots also have the ability,
there we go, to kick off cloud agents.
And so when it comes to the actual engineering work,
what each of these IC engineers are doing
is they're creating agents in the cloud
that sort of run with a local copy
of your code repository.
So when it comes time to make changes to the code base
or maybe they need to install some set up scripts
to be able to ensure that the repo is testable and runnable,
these agents come with all of those batteries included too.
So in this case, it looks like you just said yes
when Nova asked about creating a PR, is that right?
That's right, yeah.
And so Nova has gone off, launched this cloud agent,
will continue to monitor its work when that completes.
You can also have an additional verification loop.
The engineer can review the work of the cloud agent,
can hand it back to either a QA team
or your engineering manager
who can actually run an additional layer of checks
and verification to ensure that the outputs align
with the intended goals that you gave your team to start.
Yeah, I think what this is really cool about this too
is you can sort of choose how much human
you want to bring into the loop, right?
So maybe for some of these cases,
if it's a really easy change to say a documentation website
or maybe it's resizing something in our mobile app,
it feels like a pretty low stakes change
and sometimes we can just let our agents run with it
but perhaps there's a more meaningful change
if we want to make sure that this design was implemented
the way that we spec'd it out
or the way that Pixel helped us spec'd it out,
we might want to jump in and ask for a demo or prototype.
And so one thing we found that's just generally true
when working with agents is that
giving them an environment to verify their own output
in makes a huge difference
but also giving them the tools to help you
as the human verify the output
also makes a huge difference in outcomes
and you can kind of work with those things together
and get better results, so.
Amazing, so I think our engineering team
is gonna cook on this in the background.
And so what have we seen so far?
We started with this open question about the business.
We started talking to a data science colleague, Ashley,
to help us better understand some of the funnel dynamics
where the biggest fall-off was.
Handed all of that context over to a PMP agent
who could actually help write the product requirements doc.
That was the first pass.
We then handed it off to our designer
who created a first version of the mocks
to actually go and implement.
And then Emily and our engineering team
are off now cooking it with our cloud agents.
Really cool system and again,
you can really adapt this to whatever workflow
makes the most sense to your business.
And so we've seen a lot of folks who wanna go really deep
on the requirements phase
and maybe work on really complex systems
where there's lots of knock-on implications
for every small requirement.
You could imagine having a team of folks whose job it is
to make sure there's a knowledge base
that's cupped up to date
and that those requirements are fed into the product.
Or maybe you have a really convoluted code base
with a bunch of different repos that need to stay in sync.
That's another workflow that you could build in
with Grockbot.
And so my point with that is just stating
that this is one example of how you might set up
your team, but really the tool is meant to be flexible
for you to kind of adapt to whatever workflow works
for your own organization.
Cool.
Okay, now a few of the hard-learned lessons
working with Grockbot, a few things to keep in mind
as you start to build out your own teams in Grockbot.
First is this idea of named agents having separate memory,
being able to learn on the job
and really thinking about these as colleagues.
They're not going to be great on day zero.
There is an onboarding phase where you can give them skills,
you can give them context, you can teach them things.
And so over time, they actually build up that context
and intuition and can take on more and more work,
which is cool to see.
The other is really thinking about sort of reducing noise
in the system as PMs.
Like there is often a lot of pings,
a lot of emails, a lot of slacks.
One of the pro tips that we've learned is
when you set up a routine to ask your agent,
if this is essentially a no operation,
if there's nothing important or urgent for me to see,
you can actually just take the work on yourself
or you don't need to update me.
And so there is a bit of noise reduction in the system.
Agents R and Grockbot is currently grooming my inbox
and will only tell me about the most important emails there
and can escalate things, but generally,
I'm trying to reduce the noise in the system
so I can keep focused on what is most important
for the team.
And then agents all the way down,
this idea that you can have managers of agents,
agents are coordinating with cloud agents.
And Grockbot is a very helpful agent system
for helping you take on some of the most ambitious work
and accelerate your product teams.
Yeah.
Very cool.
So that was a very fun demo.
That was great being able to actually make some impact
at Fly Low Airlines.
We are going to wrap there to those
who are joining us virtually.
Check that with Matt, Lauren and Roshan.
They're running back up there.
Who are currently building a startup with Grockbot
over the course of these three days,
which is very exciting.
How's progress going?
It's going.
We have a really fun business idea.
And so if you're interested in following along,
we'll be streaming for the next 48 plus hours.
So, it should be great.
Very good.
Thank you all so much.
Very good.
And for those in the audience,
we are now going to open it up to any...

questions that are on your mind. Before we do that, one quick thing we wanted to
plug is we are running a competition. You could win a trip to Starbase, Texas to go
watch a launch of Starship. It's pretty cool, incredibly cool if you ask me. If you
show us how you've integrated the GROC bot into your work, so follow at GROC and
at bot on X. And there's a post on the at GROC account right now that explains
how to and when. The runners up here are also going to win a tour I
believe to Hawthorne to go and see how some of the stuff that helps launch stuff
into space is made. And so if you're interested in any role that that QR
code should help you get started or you can just go to at GROC on X. That's our
prediction. Very cool. Are we allowed to enter or no? I don't know. I hope so.
Alright, there we go. Questions? Oh, I will keep that up so people can scan it.
We could repeat the questions up here too if it's easier. You can just
shout it out and we'll repeat it. Yeah. Oh, put up the QR code? Oh, get out of
the way for the QR code? Yeah. We also want to hear your question. Yeah, so I think
the question is about what we've seen in terms of bots managing context, how it
might overlap across teams or roles. Is that is that the right idea? Yeah. So I
think at the end of the day these act as sort of self-contained agents that
you can you can steer and control however you want. So if I understand
your question correctly, the way that these agents these bots are
architected is that each bot has its own memory. But there's also a shared memory
pool that bots can contribute to in the cases when they and you decide that
there's something you should remember for your team. And so there's a bit of
prompting we use to kind of under the hood to tune how the bots choose one
or the other. But in general, bots have a sort of individual memory pool
that they write memories to and then refine memories from over time. And so
one of the reasons that we think we really like the role-based version of
this is that you can get really strong memories for how to become a really
good engineer over time. But then you know we're also still experimenting with this
a little bit. The group chats interface that you saw really allows you to kind of
bring those things together. So you know there are multiple approaches. For
example, you could have a single builder bot that is end product and design
all in one. It can also work. It's just that we find for more complex use
cases separating the roles how it can be helpful. I'm not sure if that was
your question. Yeah I think one of the things which is interesting is like
we've been trained with early agents to worry about context pollution and
context management. And there's like a little part of everybody's brain that is
thinking about like what does agent know? What do they not know? Do I need to
compact? And at what point am I going to run out of context window? And I
think the ultimate goal is to make this incredibly simple and incredibly
accessible for everybody to use and not have to spend a bunch of like
metal work thinking about what we think it's about. And we're back. So
hopefully you all enjoyed that session on Grockbot for product managers. We're
back in the studio and we have a guest with us. And this is Eric. Eric is the
co-founder and CEO of Carrot Financial. Banking credit and payments built for
creators and growing businesses. For Carrot, I believe you're a product
manager at Instagram. Yeah I was a product manager at Joint Meta Instagram
2016. Awesome. And another life. Awesome. Welcome. Excited to have you. So
Lauren and I have been at work here during the break. We were just
talking about like exactly what we're building, what the direction is and I
feel like we're still kind of working on exactly what we want to build and
like what the software looks like. And I think first it's an interesting kind
of lesson in how the hardest part of building is actually aligning on an
idea. We're just talking about that. And so I think what we're gonna do
since we have Eric's time and he's built the business himself and he's
you know working sort of these things. Maybe we would workshop exactly what
we're building. Refining the ideal so we know exactly where we're headed.
I kind of love that because people my mom would always ask like what does a
product manager even do? And I told her it's always comes down to three things.
Number one is figure out what to build. Yeah. Number two is how to build it.
Right. And number three is actually coordinating building it. Yeah. And I
think people don't realize like AI makes the third part really easy but
like parts one and two that's still like human brains. Right. And so when
you're like what are we even doing? I'm like yeah this is perfect because I
also want to know. Right. So I think my first question so I tuned and I
cheated a little. I had some of the live stream up what you're doing at
the computer in the very beginning. So I heard it was like a website for
pop-ups and like is this for restaurant owners to do pop-ups? I'm trying to get
a sense of like who are we doing this for? What's the problem we're trying to
solve? Yeah. For me I imagine like our main users are people who want to run
a pop-up. Yeah. So that you could be a restaurant here. You could be you
know you could have your own Shopify store but you want this like a
pop-up to me like the whole Venice or the value proposition of a pop-up store is
that it's for one it's like it's in person. Yes, IRL. It's IRL. And the other
part is it's there's this almost like a limited supply factor where you know
it's like there's a time there's a time constraint. There's like you know
pop-ups. You're gonna miss it. You gotta go. So you have to go. There's like this you
know you there's a desire to want to show up and then maybe when you show up
there's a chance to participate in a once in a lifetime or whatever you know
experience or buy products or merchandise that are only accessible
through the pop-up. So I feel like there's something interesting there
where you know actually it's pretty hard to set something like this up
yourself and so we want it where we're thinking of one actually creating a pop-up
of our own. Like a rock-pop pop-up. Yeah like yeah because I personally love
rock would love some rock-pop merch believe it or not I don't have any
pop-up merch. Maybe some tater merch as well. Sure yeah like a potato in the
or I don't know a rock-pop in the shape of potato or something. Yeah. And I
think that by actually trying to start this pop-up ourselves we will build tools
that will kind of put you know shape the platform that we want to go as well where
like initially the only user of this platform is gonna be ourselves. Yeah. We're
gonna use it to make the plan and organize our pop-up store and then
we'll figure out how to actually make it a little bit more generic make it
accessible for other users to actually use reuse the same intro to do bookings
like finding venues you know planning invitations and things like that. Okay
so this this makes sense so I have one thing I really like this and then I have
one push I want to give. Yeah. The thing I really like is especially in a
world where AI is eating everything IRL matters even more like the whole
point of what we're doing over the next few days like you can build an
entire business with rock-pop but the special part is to your point
getting to meet and interact with the business in real time if anything what
we're describing now is software it's happening in brick and mortar businesses
to like if you're a restaurant the big question you have to think about is how
do I avoid just me commoditize and put on as another icon on Uber Eats and
DoorDash like how do I make sure I'm not just a cloud kitchen that people
just ordering burgers from pop-up or like I'm selling stickers online and
everyone's just putting out AI generated stickers how do I make sure
people actually know what I represent stand for pop-up so like my first
thought is like I really like this because you're trying to recapture the
essence of like here's something special about your brand and how you can
reach people in a way that's become really valuable today my second and
here's a bit of where my push is I really think it matters who we're
doing this for and I'm hearing you describe the initial target demo for
we're building it for you said let's build it for ourselves so if we're
gonna build it for ourselves then the framing is almost people who wanted
pop-up that have never done a pop-up before and barely understand the
business side and that's okay because then I heard you say but okay we're gonna
later on go more advanced because there's a totally different demographic
where I already run a restaurant or a merch shop and it's working but I want
to run a pop-up to get attention right I think those are two really
different demographics I'm super down to start with the first one my only
ask is then let's like narrow it a little so we're gonna start with like
people haven't done it before you mentioned like merch so are we gonna do
we doing this is like a merch like hey you right this is so met up if you're in
San Francisco it's like you run a venture-back startup and to stand out
against all the other AI companies you want to have your own IRL merch drop you
want to have your own plushies like very meta is that gonna be the
initially case we built for I think so that could be interesting it feels a
bit meta it feels a bit disconnected from just like maybe is there like a
simpler way to like connect with our audience like I think if I'm
thinking about myself like what I want I don't know what a pop-up is but most of
the time what I'm looking to do on like a Saturday is just go find something fun
to do they go with a friend of mine and like I like art I like checking out
new things if there's something cool to buy like buying these cool things so
yeah I don't know my thinking is like always simplify but I'm like I'm down
for whatever you know so I mean I think we have we have I'm curious
to chat things like this is you know a fork in the road we can either go a
bit meta and really build for ourselves which is we're a tech AI brand and we
want to get more of an IRL connection to your folks so like let's go and do
merch or something or we can do something totally different like kind of
like what you said right so I'm curious is there anything in chat also these
messages are so funny the poor AV team am I just like way too loud right now
the sound guys are having a lot of fun right now bro is proclaiming
holy at least I feel like my problem is like I'm always louder than
everybody else and like when I do podcasts I just like amp down my
sensitivity a little bit okay chat let us know like our which direction should
we go or or is there just a lot of lag going on right now
no very I think there's a bit of a lag between the but I like where we're at
I think we have a good direction so like what would help most like maybe if we
start wireframing or walking through user flows we can talk about it I think
we got to be specific though on which one we're gonna do like are we gonna
be like hey like let's do tech brand doing a pop-up for your brand or should
we do something to your point like very brick and mortar very different
I'm a restaurant owner and I want to do a pop-up to get people to come to my
restaurant it's like a special event I do think we do we got to make a call
otherwise we're gonna build like the most generic pop-up say possible which I
I fear is that it's too generic to be helpful to anybody yeah I think I
agree with that I mean like what are the constants right the constants across
all of these things and what we want to do are that like we need a date
we need a time we need a venue we need to collect a bunch of different
people who have a trade into that same place and so Lauren started working on
some prototyping tools for us to start doing those things and then what is
specific to us sounds like the merchandise might be you know what let's do
like what every startup has always pitched to every VC ever guys we're
gonna start small the addressable market is everybody yeah maybe we start with
merch for tech companies because like hey we want to have an irl presence
but once we start to build it and figure out the primitives then we can
expand it to like maybe other sort of pop-up scenarios what do we think of
that I would address chat but chat is like on a
three-minute delay so we're just we're just gonna fly blind and say all right
you know what we're we're gonna try and start a grok pop-up
and sell what grok push-ups tater plushies sure
crock hats crock hats crock hats crock okay well this is this is the easiest
thing possible if I put on my product manager hat right it's like you know
the classic frame is like what are the problems we have I think the first
one is just like if we're really gonna build it for ourselves
what do we sell yeah where do we sell it coordinate the location time so it's
what do we sell number one number two where do we sell it number three coordinate
the location and time uh the fourth thing I would say is like ticketing
where you want to make sure you have a social element I think part of
full's biggest success with you knowing that all your other friends are coming
as well yeah I like that and so like we want to make this an event
like we can't have Elon tweeting out every single time the pop-up is open
right I mean that'd be huge that'd be amazing right new business to have so I
think the first question is like what do we what do we what do we want to make here
cool yeah so maybe we start with um what do we want to sell
merch sounds like we're doing a merch pop-up uh sounds like we need some sort
of system for like ticketing or like just basically like an event
registration we can probably find like an api or something that like helps us
with that um or we could just use our our plan scale database we could uh
do some like quick event registration make it social so ticketing
and we don't even really need to worry about time slots like if we wanted to
like do it across the day or something right now so um merch we're going to
work on the merch for the pop-up um that should be integrated into the platform
um the registrations I think uh like that seems like pretty easy and then
maybe something to help us like finding the venue yeah like a venue yeah yeah so
maybe you I love it you simplify to three things just the merch the
ticketing with like the social piece and then just finding a venue
to your totally right time slot sound is important page one what are you
building but in this instance it's you're building a merch pop-up page two
um is literally like what date is this thing uh and then it helps you like build
an event um a booking system and then page three is like uh uh kind of like
where are you where are you having it and then you can you can start to to
to figure out the location and whatnot so here's my question is someone who's
still learning grok bot what's the recommended approach using grok bot
to figure it out like are you going to have each of those like merch
ticketing venue is that a different bot or are you going to split it functionally
right in terms of like engineering versus design like how how are you thinking
about this lauren or lauren is currently i'm like super long maybe maybe
not how do we split this up i like to split these things up i think about
bots as like different employees really so like in my bots right now i'm
i set up on my screen i have a growth engineer i have a founding engineer i
have an image generator bot i have a creative director chief of staff and
then some of these other things which usually come from like tasks like
if i have a task for example updating my documentation i'm going to create a bot
separately to update the documentation and so you can see here lauren actually
has been just cooking on some like yeah i'm so yeah what i was doing in the
background and and i was like kind of zoning out of it but
or not zoning out about trying to lock in on this was to start so i'm a very
visual person yeah and so i got my prototyper bot to
start stretching out stuff for me in tl draw which is pretty cool that the
so one thing grok bot does really well is like you can actually control
a browser and it actually when is the tl draw is that a drawing for me
all right so i we have some kind of very low fidelity mock-ups that we can kind
of just critique and think help us ideate what we're trying to do
so i've just been prompting and trying to get it to work on that one one
thing i'll say i saw so many in chat say merge just merge you know if we're
going to put on our true like silicon valley satire hats i would say this is a
community as a service and a world dominated by ai we want to help you figure out how to
bring the in-person vibes of your company back so i think if it less is
merge this is a platform for you to have community meetups around your brand and
yeah as a tech brand one thing that's really easy is have a reason
merge is a great reason sweet yeah totally good and i know uh i know some some
providers like i know some apis where we can maybe line up some merge um that might we
might be able to integrate into the platform some ways we might be able to make
that happen so sweet well well when if you're working on that i'm not sure i
think you are working on that i guess i'll ask the other person yeah no i'll
i'll clear up some time lauren do you feel like we have what we need to start
building stuff uh i'll try okay cool yeah i realized the two of you are like
hosting and guesting at the same time it's pretty hard to like
incredibly difficult to be entertaining actually build well then i'll ask you
so like you know if we're gonna find it's like hey if you're a brand
you're gonna have like a community pop up reusing merge is like the trojan
horse to like something to start with what's like fun merge because i feel like
every tech startup ever has done i'm we i'm wearing merch from my own company like
t-shirts and hats like is there something unique to grok i feel like we could
lean in the direction of cute like taters and potatoes because that's part of
freaking her brand because there's something something different i mean
obviously you know how to the flamethrowers like that is a unique
type of merch if you think about it yeah so we just have to figure out
flamethrower production in like 48 hours sounds like uh no i mean i think
when i think about the best merchants similar to uh what you described was just
like building for yourself it's like hey yeah i dig hats but i did very specific
kinds of hats like i like hoodies but like i want a very good specific kind of
hoodie um and so if you can like work backwards from that like in the
clothing aspect um it's kind of like what kind of stuff do you do like one of
my good friends gave me like a snorlax plushie at one point are you are
you pokemon fan i mean you know i have friends that are pokemon fans
okay i play a great why and i love that and it was like very
reminisce of my childhood you know it's kind of like cool gift
um and something like that resonated with me so it's like how do you find
something um either sentimental or emotional that you can connect with
some of you so my my two thoughts here the first one is like that could be a
bot action itself as we're exploring use cases until like go in research and
figure out what people want my second thought is basically you said i heard
two things i heard one an expressed appreciation for utility like not just a
hat a really good hat yeah the second is a desire for novelty like the
snorlax thing this is a total tangent on the novelty piece this is not a utility
piece have you heard like uh i saw this tick tock the other day it was like you
know boys get girls flowers not to be gender normative sometimes people in
a relationship like to get the other person relationship flowers what's the
other version of that right like a girl was like what do i get for my boyfriend
and somebody suggested and tick tock tungsten cubes
tungsten have you seen these love tungsten cubes yeah why they're
they're so they're because they're cool they're so they're so happy
they're really dense they're really hard to pick up so i'm like chat's gonna
lose i'm not saying we're do well do we do this but if we're gonna go the novelty
direction like in addition to the like the hats and i don't like tungsten cubes
yeah 100 yeah i mean i don't know why we're i love he immediately understood
what i'm just telling you i don't know can you get like a tungsten cube in the
shape of a brah-pot you could get a that would i mean a spherical
that would be crazy you can absolutely do a hard to tungsten
tungsten god wait that i would actually want guys like i'm really
hypothetical i would want one yeah i mean how much would you pay for it i would pay
honestly i pay 80 bucks for that 80 bucks and here's the whole thing it's so heavy
you can't even get it shipped i mean you have to go to the pop-up to get it
back off a little bit of one inch one inch tungsten sphere is $500 so uh $80
happy hard now tungsten cubes maybe with the grok bot logo on them are less
less expensive but the tungsten sphere is $500 so 500 that that's an expensive piece of merge
okay so we might have to pass on the this might be primo merch maybe we have like three of these
yeah people at the pop-up or something maybe one person okay i'm going totally off track maybe one
person out of every 500 that goes there wins the tungsten sphere i hear you i hear you yeah i
can tell you like that maybe we start uh lauren with like the basics like we can back into uh
where we're booking what we're booking some of these other things i know your cloud
agents are working on it in the meantime i've also been working on um some of our like site
stuff just connecting email infrastructure i did want to take a moment and talk through like chat
because we've been getting some really great messages um about what people like what they
don't like we have these in notion um so from the joy t if friends can use each other's bots
do they inherit the bots previous training can you remove past behaviors retrain or reset it
and then refine it in your own version so it's interesting currently the way these bots work
like if we if lauren just sent me a bot that would be a template um that template basically allows
you to copy the memories um some of the context or what the bots worked on or like the instructions
as well as like first party plugins and some other things without carrying out of
our sensitive data or like chat data so if you shared a template with me you wouldn't have
to worry about sending me any passwords or number one the bot can't see passwords we'll talk
about a little bit more about that in a second but any sort of chat information that's specific to you
related bots store passwords securely using like secure entries so you actually don't need to worry
about bot access to any sensitive data you'll get like a nice little form um and so basically the
the answer to that is when you get a bot template you're going to be able to refine it uh as you
go it's going to have a certain set of behaviors and actions and it's necessarily like your own
kind of version of that bot it's not an exact one for one copy um i'll talk through a couple
of these while lauren's kind of iterating on these designs um interesting to see her uh her process here
how can you tell if you've given your bot too much scope what errors do you observe over time
from crystal and then a very long string of numbers um i think you just have to observe
kind of how they respond you know the parallel here i use a lot of parallels to like humans
and it's like if you ask a human to do too much maybe you're seeing some of this on the stream
uh the human starts to get disoriented and might say something's like they're a little flustered
or they might not make sense and so you can start to observe in ai like um is this making sense
are the outputs high quality um and if we think more about like why that might be how
context work how next token prediction works uh it can start to make sense but we won't go too
deep there i would just say does it feel like the bot is on target does does um your ai feel like it
is moving in a good direction um and the if the answer is no you might need to narrow the scope
for clearly context the answer is yes um then maybe you're feeling better uh so interestingly i'm
noticing lauren's bot here is like yeah my bot is really like literally sketching

Rock Barret Bot Art Exhibition.
I prompted it earlier that we might want to host an art exhibition.
I forgot to mention it.
But my idea right now is I want to have, again, a very visual person.
I actually want to come up with a cool ticket design.
Because I feel like that drives some kind of...
Oh, 100%.
It gives us some momentum.
Like, oh, we got a really cool ticket.
I just want to, this creator event, it's hosted by Make
creating William Osmond.
And you're bad just to get in actually PCB boards.
Like, you can actually wire them up and have them light up and everything.
So, no, I love this.
You come up with a cool design for this?
Yeah, right now I'm just thinking like, well, I'm sort of inspired a little
by like Pokemon cards and how they're a little bit shiny.
Oh, wait, so you want, in some ways, the tickets to the first ever
Rock Barret Bot pop up are collectibles in themselves.
I mean, that could be pretty interesting.
That is really cool.
Yeah, yeah.
Yeah.
So you already have...
So the ticket itself is a collectible.
Yeah.
And then you show up and then you have a chance to actually look around
and buy more cool merch.
Yeah, because like, you know, keep in mind, the whole reason we're doing
this pop up is because it's special.
It's not just software.
So I feel like having a really cool ticket, that's part of it.
So my question, again, is somebody who's still learning about Rock Barret.
You have one right now designing the tickets themselves.
Are we able to have one like figure out like the merch?
Like tell it, you know, hey, like we don't just want plushies and high quality hoodies.
Like, we want that.
We want that for sure.
But we want something cool and novel.
Like, hey, we consider like the tungsten cube.
And then we consider it the tungsten cube in the shape of a
crockpot and we realized that would cost like $3,000.
And so does it have other ideas in this vein?
Yeah, let me just write that down.
So we have I'm curious if you have more cool ideas like tungsten.
Is that how you spell it?
Yeah, that's how you spell tungsten cubes, which again, he and I instantly
in the old usage of tungsten cubes.
Give me a throw out more ideas that I can try to.
Another good novel to you once.
So playing through.
Some of what we're seeing in the chat, potato plushies.
A lot of people seem to like potato plushies.
Pictures from SpaceX launches, so rocketry pictures.
Oh, OK.
Pictures from SpaceX.
Limited edition crockpot stress balls shaped in the different
shapes of crockpots.
Oh, that's that's a really cool idea.
I like the stress ball one.
Probably not too hard to make.
What could be cool is like, you know, how rock
bot is spherical, right?
Like a hat, but the hat just has eyes on it.
Like the hat is in the bog.
That's cute.
Like to your head or like a toboggan.
Oh, actually, this is you probably just heard me say
something that sounds like a sled.
This is I'm from North Carolina.
Yeah, like, you know, like a beanie, like a beanie thing.
People in the South call those toboggins just completely
sidebar on the trailer.
I think it's a lot.
Yeah, I thought it was really crazy.
So I just had to go on a tangent there.
But anyway, basically like a wool cap of some sort.
Right. So another one I'll add.
Wait, what about rock bot character plushies?
Like the whole point is you have like a whole army
of crockpots.
We can make like cool little plushies for each of them, you know?
I also feel like it'd be really cool to have the attendees
of this pop up be able to like participate
by like, I don't know, submitting some ideas
and then we'll like we'll make them because like in rock
bot, you can create your own bot templates.
So I wonder if there's also something we can tie in there
where, you know, you're you can submit a bot template
like we'll choose good ones and then we'll
make some special merch for them, I guess.
I don't like wait, can I go one step beyond that
for something even crazier?
And I was thinking about it.
Right. So what if you could create your own physical
grok bot that also has its own purpose
because you can code it, right?
Like for example, you could make a physical grok
but I'm just making this up because I'm in a long
distance relationship right now.
We have find my iPhone turned on.
What if I like got a key chain and the grok
bots that bots agents soul's job is to track
the location of my partner.
So I know where she is.
Or like you've seen this in Tiktok.
It's like she's X miles away.
So I'm not saying you use as an example.
I'm saying what if you have people create physical
grok bots with like specific things or like the
other classic grok bot I've heard used is like, oh,
if you're a startup founder, it gives you like the
competitive news update of what's happening in your space.
What if you had like a physical one who's looked at it
and just tells you like, this is what happened.
I see. Yeah.
Or maybe it like tracks sales of your merch store
and every time, you know, you make like a thousand
dollars, so lights up.
It's like you just made another thousand dollars.
I don't know, because it's like kind of taps into the
ethos of you can customize it and you can code it and we can make it.
I'm really just thinking out loud now.
Yeah.
Yeah.
Well, this is someone commented, grok bot,
build a bear kind of thing.
Yeah.
Another person commented, don't do that.
There are two types of people on the internet.
Oh, are they all right?
So this one's still working on the tickets, right?
So are we just not going to do the platform idea then?
Well, I figure like we could, if we build the tools to do it ourselves,
we can then make them scalable.
Yeah, I feel like the platform maybe is like a stretch goal.
Yeah.
All right, we'll build the tools for ourselves first.
I like that.
I think we, like, I think we should focus on just getting the store.
I fully agree.
Design, setup.
The sort of design, finding a place to do this.
Exactly.
As soon as possible.
Getting all that done.
Tomorrow and the third day we have time to actually go
platformize things if we feel like they're worth doing that.
But for now, yeah, and then I have a merch box set up already.
And it's coming up with ideas.
Build-a-bear, Grockbot thing, if it's not impossible to execute.
And then in the course of doing it, yeah, we can extend it later
to let anybody do their own pop-ups.
Oh, actually, I'm going to say ask Grop to use Grock, imagine.
That's cute.
Tom-a-Grock, a Grock Tom-a-Gotchi.
Maybe that's what the story has, is it a Build-a-bear?
Maybe it's just the Grockbot Tom-a-Gotchi.
Oh, yeah, we're telling Steve, like, hey, don't just regurgitate.
Yeah, don't just tell me what I just told you.
Come up with something more and better.
It really does feel like you're just messaging a team of your colleagues.
All right, I think one of our colleagues is actually back.
Hey!
How's it going?
Good to see you.
It's a pleasure.
My name's Eric.
Hey, Roshan.
Nice to meet you.
Of course.
I caught some of the conversation, but what I missed.
It's got an interesting direction.
So the thinking was, look, if we do a pop-up, we should have a clear sense of who we're
building it for and what we're trying to solve.
Right?
And I think there's something very special about getting you an IRL manifestation
of what your brand is, especially like with AI software everywhere.
And then we're like, this is actually Lawrence's idea.
Why don't we do it for ourselves first?
Yeah.
And then they go, well, for ourselves, we can expand this into a platform.
So then we started thinking through, and this is actually Matt's point,
we're going to do it for ourselves early.
There are only three things we need to figure out.
Hold on, hold on.
Let me, maybe I should record this.
Should we like, feed this into a bot?
Have we done that yet?
No, it's, and we can update our docs.
I have a, I have a outreach bot.
Now we have one working on menus.
Let's find my outreach bot.
Okay, here we go.
Sorry, pick back up where you left off.
Yeah, go for it.
And so Matt had actually a really good point.
If we're going to do our own rock-bop pop-up, there's really three things
we need to figure out.
Number one is just what's going to be there, like the merch.
The second thing is the venue, right?
And the third thing is ticketing.
And again, all of these can later be extensible into like their own
broader sort of pop-up things.
So we started working and thinking about like the merch and the ticketing.
So she has one rock-bot actually designing the ticketing, which in
itself could be a collectible item.
Yeah.
And we have another one thinking through merch ideas.
And Matt expressed that from a merch POV, there are two things he
cares about.
The first is he like utility and craft.
So like a hoodie or like a hat, but like a really well done one.
The second one is like novelty.
Exactly.
The second is novelty.
So you talked about how a friend once gave him like a Snorlax plushie.
He really appreciated that.
And I mentioned, you know, as a guy, I feel like my favorite
novel tour is like a tungsten cube.
And then Lauren Brown's some ideas actually where it was like, oh,
maybe like, it's not just a tungsten cube.
Maybe it's a tungsten figurine in the shape of a rock-bot, which
then evolved into what if you could actually customize it?
Like what if you had different rock-bot plushies because there's
so many characters, which then evolved into what if it's going
to be a plushie?
Why not program in some functionality into it as well?
Like you could have one rock-bot plushie.
The job is just to update you.
What's happening in your slack?
That's cool.
Honestly, I don't know if I would actually want that as a
founder. I feel like I'm on my phone checking it all the time
anyways.
But there's something interesting here on like, if we're
talking about pop-ups, it's the manifestation of your company.
The rock-bot plushie could be the manifestation of a rock-bot,
right?
Where you can actually program it to do simple things.
And that's where we left off Don.
And then we were thinking more about how do we actually
leverage the bots to do this?
Yeah.
Like one bot's thing you're ticketing, one bot's
thing you're the merch.
And then we got very lucky and you came back.
So that's that's sort of catching you up.
I just recorded all that.
Yeah, of course.
So let me maybe feed that into a bot.
Maybe we spin up a merch bot whose job it is to help us
like, yeah, paint a bunch of really cool merch and
figure out what's doable.
Let's do it.
Cool.
Okay, cool.
I got my merch bot hooked up to Grockimagine, but I
guess it doesn't really know.
It's got some really random.
Which I wish it kind of cool, I guess, but it's not.
It's kind of messed up the Grockbot logo.
I forgot to tell it to do that first.
So let me try and cook more.
Maybe we should also like, would it help if we found
a supplier and then we like grounded some of this
merchant things that actually exist?
Oh, yeah.
Or like we had a way to basically like maybe feed via
API or website, like what merch options it has.
And then Grockman like picked that merch.
You could like go to, you got like scrape Alibaba, right?
Or like Tabao.
You could even have it like do outreach to like Alibaba
and Tabao and spend like 1000.
We're probably not going to do this.
I'm just saying it.
Tungsten Q manufacturers and be like, how cheap can you
make them?
I'll start doing some research on how we can do that.
Using Grockbot.
I feel like we should come up with like two or three
really just like perfect examples of merch, you know,
like two or three really well curated things that we think
we could actually get.
For sure.
That might be a good way to like pare it down.
Totally agree.
Yeah.
He's like, there are so many ideas.
He's being a good PM.
He's like, we got to only do a few.
Yeah.
Yeah.
That's part of it, right?
It's like you want it to be like intentional,
like sort of intentionally thought of, intentionally designed.
And I don't know about you, but I feel like sometimes
there's this bit of like less is more where you really
want to, you know, like knock it out of the park
on one or two or three key things.
So.
Well, then I'm going to toss the PM hat to you.
We've discussed many, many ideas.
What do you think is the first one?
But you're most excited.
For the merch?
Yeah.
Or for the products or everything we just mentioned.
Well, I don't want to scope creep too much,
but I think it could be really interesting when,
you know, maybe if you're registering for the pop up,
you can choose your favorite merch item.
Like maybe we could show you a gallery of some of the
possible like we could make a little bit of like a
gallery where like in the registration experience,
you can like, you know, pick your merch item.
You can kind of see some previews of what that would look like.
Maybe like something cool.
This might be a stretch goal, like a 3D interactive thing.
You can spin it around and see it in 3D space.
I know that the, it's like a bunch of MCPs out there
for like 3D design and stuff.
That could be a really fun thing to do.
Or I feel like one of the things that I always think
about when I go to events is just like wanting to know
that I'm sort of, you know, getting something unique
or getting something special.
And earlier today, we were talking a lot about how
we want this to feel we each want to bring something
special to the event.
So I wonder if we build that into the registration experience
and that could actually be something.
You know, if we go in that direction where we get to
leverage like the graphic customization capabilities
and like show visualizations, it really makes me feel like
we should just lean more into like for this IRL pop up,
build a grock bot.
Like you have characters for your bots.
You can like create a physical manifestation.
You can design it.
You can visualize it.
You can see it and you can assign it a function.
Right.
Like kind of what I talked about.
It doesn't have to be slack updates.
Maybe it's just like the weather outside.
Although technically if you're outside touching grass
already, the fact that you would need
your grock, plus you to tell you what the weather is,
is a little to soak in.
I could always use an assist.
I don't know.
Yeah, you're not sure.
You're just like, please explain it.
Hope you interact with the world around me.
I like that.
But like, but giving people ability to customize
literally their own IRL grock bot, like that, that
could be fun and you could visualize it using the tools
you said.
Yeah.
What are your top three?
I heard a bunch of ideas.
I think I mean, to be honest, after hearing you
describe this visualization idea, I think my, my
favorite one is this fact build, which again,
was somebody from chat, like build a grock bot.
Yeah.
It's like this entire workshop is us building your own
grock bots to build a company that will now build
you physical grock bots.
Yeah.
Yeah.
Yeah.
Very meta.
It is very meta, but it's like, in a way it's a loop.
Yeah.
A strange loop.
Yeah.
I love that.
I love that.
Um, awesome.
Well, it sounds like we have some merch ideas
floating around.
Uh, well, yeah, I have a few.
I mean, I need to iterate a little bit, but like,
Oh wait, wait, that's cool.
Those could be some, I mean the eyes are a little
bit off that.
They're just very tired.
I don't know why you're like, like that, but we'll
just need to tweak it a little bit.
That was a cool, cool comment in the chat.
And then I also have another bot working on
creating some cool ticket ideas.
So you can actually like, well, I'll probably want to
find a printer.
We can actually print it would be good.
I know I know some merch connections.
So we could start to research that.
I can start doing some research on like a printer
for tickets using.
Yeah, tickets.
Okay.
Oh, I think that tickets are almost like a
collectible and stuff.
And then like you sign up and then we'll like
meal it to you or something.
Well, we give it to you on site.
We give it to you on site.
I'm sorry.
Coming with a generic ticket.
Like you come in, you actually exchange it
for like your ticket that you can carry.
Or maybe you have a badge or a name tag.
Now we're going a little crazy.
No, it might be worth downscoping a little bit
and just focusing again on like the system that
we're building and like what we're trying to deliver,
which is an experience for which we need a venue,
a merch and then the people that we're going to collect.
This is a good merch exploration.
But are you really locking in the venues?
I mean, we need to work on that.
Wait, let's get a bot to like figure out the venues.
I have a list of venues that I know about.
Should I make them a bot?
Do you have a bot template you can share with me?
A bot template for what?
Finding venues.
Five vending venues.
I'm sharing my screen, but I guess I should show.
Yeah, absolutely.
Let me find the venue, my operator research here.
Create a template and I'll get that over to you.
To me, the best part of every hackathon
is when you get to a point where like,
wait, I actually want this to happen.
Like I actually want there to now be this pop up
where I'm going to go and get my Grogpaw plushie.
Yeah, I love that.
You know, and maybe you can actually do things as well.
Yeah, yeah.
This is awesome.
Yeah, OK.
Wait, so who's going to figure out the venues?
I know you already haven't pulled up,
but you have the screen share.
So which one should we go with?
Oh, the venues are in Notion, actually.
Oh, great.
Well, maybe we should have a Grogpaw.
Just like reach out to all of them.
You could just be like, where's the Notion?
The Notion, what do you mean?
The link me the Notion.
OK, sweet.
I'll send you a slack.
To the comments, I am on X.
I'm really bad at X.
I have very few followers, but if you just Google,
if you Google it, you'll find it.
This is my name.
Some of you were asking.
I don't know about what are these?
OK.
No, I'll say.
Yeah, a little more interesting.
But it's quite like.
I think it doesn't really know how to.
Yeah, I like how the bot just threw these up
on a local URL on its computer.
Oh, is it local?
You could just go and yeah, and just check them out.
I think it's also funny that your merch drop bot is named drop.
And I asked to create a merch drop bot that also got named drop.
And so I guess.
It's too.
It's too.
I bought it.
So.
Right.
Converting evolution there.
Yeah, yeah.
This is a total sign note, but I feel like you know how you
have templates right now or for the bots you can share around.
Yeah.
I feel like one day you could have like a whole theme
where you can spin up an army of bots and they all follow one theme
like potato named puns.
Or like Sonic Sonic the hedgehog themed ones.
And just spin up more bots with those names.
All right.
This is cool.
Lorna just sent you some of the the venues.
And then I'm working on getting a little template for you.
Oh, I'll figure this one out there.
Like I want to sign in this.
Yeah.
See someone said if you genuinely sell rock pop plushies,
I would buy one.
Yes, good.
So someone said told me to get good at X.
I'm trying guys.
I'm trying.
Yeah, I really like this like build a grok bot plushie.
But like maybe you can program an actual functionality
or if that's too complicated, maybe it can do like one of three things.
You know.
Yeah.
Also, I feel like it's just good for the grok bot brand.
You can people around San Francisco in the world carrying their actual
rock bots.
I totally agree.
Yeah.
A little bag change or something.
Yeah.
Yeah.
It's not just an AI interface.
Yeah.
Yeah.
Very cool.
Very cool.
What else did you guys cover while it was gone?
So it sounded like there was a bunch of discussion on merch.
How's the rest of the building going?
Well, she's got to get any designs.
We just talked about like figuring out venues, which I assume would be
outreach to the venue owners.
That's pretty much it.
So far.
Love it.
Love it.
Awesome.
Yeah.
Cool, cool.
And we're also coming up on our next session, I believe.
Got my little papers here.
So we'll have in about five minutes, six minutes,
grok bot for founders.
Let's show one of our very own startups growth folks.
So we're going to cut over to that in about five minutes,
just to heads up.
Cool.
You know, one thing, Eric, that I wanted to just pick your brain on.
Yeah.
So I know, Carrot is like a platform for creators
and folks who are creating content, right?
Yes.
And one of the things, these guys are going to hate me for this,
because I've been like harping on platforms all morning.
Yeah.
But we've been talking a little bit about how, you know,
we're going to go run this pop up.
But maybe some of the software we build is generalizable
for other people to run their own.
Yes.
I think that's the world.
I'm just curious.
Like, yeah, what are your thoughts on just like platform
businesses and as your experience kind of being a founder of one
of these?
Yeah.
I'm curious.
Like, obviously there's a stretch goal for us coming out
of these three days.
But, you know, just given your experience,
I would love to hear kind of what some of the ideas that we've
been cooking up.
Actually, what parts of what we've built,
do you think would generalize?
It's a brilliant question.
I'll say two things.
The first thing is applying to my own company.
And the second is like, for example, one used to work
at Instagram.
So for our own company, so the one liner I gave when I came
on is we do banking and credit for creators.
Yeah.
Over time, that's actually expanded because, yes, we
initially built for YouTubers.
So some of the biggest streamers in the world,
like Alex Botez, a trust streamer, or like Ludwig,
they actually use us.
But when we built all of the primitives to help these
creative entrepreneurs bank their businesses and get
credit, at a certain point you're like,
why can't I extend this to other types of businesses too?
Yeah.
And so now we've gradually expanded to other types
of founders and small-mean businesses.
What are the things that creators like love the most
when it comes to now?
Such a unique user to be working with.
So today, banks generally suck because they're just dumb
buckets for you to put your money in and then nothing
ever happens.
The dream is you build a bank that actually helps you
grow as a business.
That's what we're releasing next year.
But for creators, an example of that is we put them
up on billboards because for a creator, when they're
featured up on a billboard, that's content for them.
Or we put together events where they meet each other
and they network.
And so for me, I think of a bank as an archaic term
we had to have in the 1950s when you needed a brick and
mortar institution holding your dollars.
And now we, in this new age, have to think about
as here's a company that's licensed and regulated to
hold your money and has federally guaranteed
insurance by holding your money and seeing the
inflows and outflows, combining that with AI,
what can it now tell you about your business
that you didn't know before that would help you grow?
That's the direction we're actually moving.
I love that.
And for those of you who are on the stream who saw
Cody Sanchez earlier, I think there's some callbacks
to that in terms of helping, really it comes down to
helping people grow their businesses.
So for creators, that's a creator business where
they're pouring their heart and soul into it and
want to be able to go to that.
For maybe folks who are coordinating events,
maybe that's an angle we should think about
with the platform here.
How do you really focus on helping them grow
their top line, their bottom line,
really just bring revenue into their own business.
For me, and I've had Cody in my podcast as well.
And I think of it as the exact purpose of what
we're doing right now is also to answer your
initial question.
Grockbaught might initially be adopted by all of us
in San Francisco who are very AI-pilled.
But the broader goal for AI to truly have a
beneficial effect on the economy is for everybody
to learn how to use it.
And what I think you've done with Grockbaught,
you've built it in a user interface and with
familiarity such that we hope one day all small
meme business owners would use it.
Like Joe Plummer might want to use that.
That's why I think it's really interesting
to your initial question.
We're building it for initially's case which is
like us.
But if we get the primitives right,
is she be able to expand to much bigger demographic?
That's what we're doing at Carat.
That's also what we're doing right here right now.
We're doing the pop-up for us,
but eventually it could be a platform for others.
Yeah, I love that.
So I think we have three minutes left.
My one thing I was curious about before we cut over
to Shubh, when it comes to creators and Grockbaught,
like what do you think for any of the creators
who are watching and listening on the stream right now,
like what are some cool ways creators could use Grockbaught?
Just from your experience kind of having put the product
a little bit and kind of having hung out with us for a bit.
So my main thought is AI is most useful,
and this actually I believe comes from CatGPT.
AI is most useful when it generates output
that is both correct and verifiable, right?
The main issue today is when it's not verifiable,
you don't know if it's good.
That's how you get quote unquote slop.
I think for creators, what's correct and verifiable
is actually less the creative part of what they do.
Yeah, it's more the monetizable piece.
So if you're a creator, brand deals, right?
Today you're traditionally once upon a time in Hollywood
to put in on agents and gatekeepers to figure that out.
Now anybody can build a brand online
and then go and monetize it by finding the right brand partners
by having four different Grockbaughts, right?
One to figure out which are the right brands to reach out to.
The second is actually go and reach out to them.
Third is to negotiate the race.
The fourth is to come up with content ideas.
Like that's just one example for a creator.
I think there are versions of that for every business owner.
You just look at your workflows and ask,
what is a workflow loop that is verifiable?
Like I can look at it and know if it works.
Like coding is an example of a loop that's verifiable.
Totally, yeah.
That's where you should go and have Grockbaught just go crazy.
I love that. I love that.
Cool. Well, with a minute left, I guess,
I know you said you're not super active on X,
but there's tons of love in the chat.
Where can people find you?
Where can people learn more about what you're doing?
On X, sadly, it's just my company, how to try,
Carat. It's like not very large.
I am so sorry.
If you Google-
It might be larger after this.
It might be. If you Google me,
you can find all the other places that I'm online and bigger,
but I would never dare say their name while we're here on X,
because that's the main platform we're talking here about.
I love it. I love it.
So I just want to say it was so much fun coming on
and like to all the folks listening,
look, you can see the excitement.
We're just like, oh my gosh,
like look what we can do with Grockbaught.
I think that's really the tick we want you to take away as well.
And yeah, this has been a lot of fun.
Unfortunately, in a few seconds, I turned into pumpkin.
It's been lovely getting to say hi.
Awesome.
Yeah. It was so nice to have you.
Yeah. Yeah.
Thank you so much.
Awesome. Next, Howard from San Francisco,
we got Grockbaught for founders.
Shubh Gaur.
Take it away.
All right. Hello, everyone.
Welcome. Really excited you all are here.
And thank you for letting me interrupt the cooking session
that's happening on the live stream
and also for joining this workshop about Grockbaught for founders.
I'm Shub and I spend a ton of time with founders
helping them optimize their Grockbaught setups and ship quickly.
And so I'm very lucky to be pushing this product to its limits.
And today, I want to share some of those learnings with you guys
and hopefully get into some live demos.
So let's get straight to it.
I'll start with what we're going to talk about.
Over the next 60 to 90 minutes,
we will run through some really quick context around Grockbaught,
why it's particularly useful for founders.
And then finally, obviously some demos.
We got to have those.
We will close out with some tips for power users
that people seem to love
and then also try to do some Q&A
for at least the folks that are here with us present live.
The final bonus piece I will add
is that I'll be sharing one of these bots at the end
so you can steal it for yourself.
I wish I had like a call now number, but we'll make do.
Also, there will be a few of you in this session
that are not founders.
You will have snuck in, which is awesome.
Kudos.
A lot of this will still apply to you,
especially if you are head of X at your company.
So stay posted.
And for those of you with Grockbaughts watching this session,
this is a time for you to probably watch it live
because we've got some cool stuff going on.
So let's get straight to it.
A few of you have probably already seen
some version of an AI maturity curve.
Everyone is freaking out about this these days.
But for us, we are really optimistic
with how the world of AI is changing.
As models improve and the tooling improves,
we are able to scale our productivity quite a bit.
And so you'll see that we started with chat bots,
moved on to ephemeral agents that finish a task
and then you kind of throw them away.
And now bots are kind of this next iteration
of what we're really pumped about,
which is bots that automate your jobs
and bots that you can have around for a really long time.
The key part is that these bots actually compound.
And I think that's the most important piece is
you let them do a task, but you don't throw them away.
You keep them around, you invest in them,
and they get better with you the same way an employee would.
Finally, what we're aiming for is this world
where you can just fully automate a staff function
for yourself and scale yourself with a team of bots.
And we're getting pretty close to that internally.
We've seen some external examples of that,
and I want to share some of that today.
But we're really excited about what Grockbot can do.
And I know this is all very high level,
so maybe we get into the specifics really quickly.
How many of you here have used Grockbot before?
Just quick check live at least.
Perfect. Most of you, it seems.
So I will blaze through this.
Grockbot is obviously a new tool that you can talk to
just like a coworker or a teammate.
You basically text it like iMessage,
and it can go off and do a lot of really cool things for you.
It is very proactive and very ambitious.
And the reason it can do those things
is because we have armed it with its own computer.
So unlike traditional agents where they don't have the MCP
or the API or some sort of direct connection
to what they need to do,
Grockbots can go off and do things on their own
even if they don't have the perfect level of access.
And we'll get into how that works in the demos really quickly.
But they learn from you.
They're proactive, like I said,
and they have their own device.
And so they can not only connect to other products directly,
but they can use your tools the same way you would.
And I'll explain what that means in a second.
But overall what you need to know is
you can assign Grockbot's tasks completely end to end.
So for you, all you need to do is delegate to them.
And the mental load of having to check on your agents
and make sure they've done their work is gone
because in the ideal state,
a bot just handles the full end to end flow for you
just like a colleague would.
And they just figure it out as I like to say.
Okay, with that, obviously a ton of founders in here.
And so you probably know what Grockbot is
and you're wondering what can this thing do for me.
And so there are a few things
that are really important to understand.
First, you all know your responsibilities as founders, right?
You have three key execution pieces
that either make or break your company.
And those are one, you need to preserve your focus.
You have a trillion things hitting you all day.
Two, you need to do that while maintaining velocity
across all of the things that your team is shipping.
And then finally, you need to really protect your ability
to make great high quality insights all day.
That means good decisions and planning for the future.
And Grockbot is uniquely positioned
to make all of this way easier.
You'll see some features up here on this slide,
but I think most importantly,
Grockbot can handle all of the distractions
in your day-to-day work.
The stuff that is necessary,
but not necessarily the most important,
you can just hand off to Grockbot.
And that mental load that it saves you
is probably more important than anything else
I'm going to talk about today.
And so it's worth you taking the time
to audit what that means for you.
The second is, like I said,
it can get its tasks done end to end.
And so that means you really want to prompt your bots
to do all of the work for you,
not just parts of it the same way you're used to today,
but really do all of it.
That includes the execution
and the verification of what it's done.
And then finally, I think there are large amounts
of information that we're all consuming every single day.
My screen time on X today is very embarrassing,
and I'm sure your guys hopefully is too.
And what that means is
there's just a constant flow of information.
In order for you to maintain
the same level of quality on your insights,
you need to find a way to triage that entire fire hose.
And Grockbott's are really well positioned to do all of that.
Okay, with that, it's pretty clear to most of us now
that this is the best time in history to build something,
especially in software.
And Lingji, who's one of our engineers,
he was talking earlier,
mentioned that he can ship
way faster with Grockbott than he ever could before.
I'm sure all of you have seen this at this point
by kicking off Cloud Agents.
There is this ability to delegate these full tasks
that I will harp on forever.
You can automate a lot of the pieces
that you have to think about constantly with routine.
So you can have them set on schedules,
on web hooks, on signals,
so that once they're taken care of once,
they can just happen forever.
And then finally,
and I think most importantly on this slide,
is you can do things that don't scale,
except you can actually scale them,
which is really cool.
I'll get into a demo around how that works,
but we've all heard the stories
of how most startups win their first customer.
There's some sort of hacky approach
where you build some very niche super...

for janky prototype and just hope that it works for them.
And then when they finally sign that contract,
you breathe a sigh of relief,
and then you do it 100 times again.
And so with this, and with Grockbot,
you can now make all of this happen
without the constant headache that comes with it.
And I'll get into a demo about that in a second.
But we're very excited about what this means
for founders specifically,
because you're all generally
the most high agency people out there.
And so if you have a product that enables that agency,
we get to some very, very cool ships
coming out in the next few years.
Okay, let's get into some use cases
that I'll be harping on today.
Obviously, you all experience your own unique pain points,
but these are a pretty good reference point
for how you can use Grockbot in your day-to-day.
And the four I'm gonna talk about
are first, closing customers.
So this is the do things that don't scale part.
It is way easier now than it has ever been
to take your customers and go with them end to end
without you having to do a lot of that handholding
and the intimate conversations.
You can really focus on just getting your bot
to handle most of it
and then step in when absolutely needed.
The second is product changes.
As you know, engineers ship way faster now,
you're probably shipping way faster
and staying across that many people,
shipping that quickly can sometimes leave you
with some pretty embarrassing moments.
There were actually times when I was demoing Grockbot
in its earliest forms
and we removed a certain feature
that was in the sidebar.
And I was so used to pointing at it.
And when it was unshipped
because I was just busy that day, I didn't see it.
And so I was kind of demoing and pointing at nothing,
which was really embarrassing.
I hope someone else has gone through that experience
because it is not fun,
but this will go through how you can stay
across all these product changes.
The third is adapting to competitors.
I think this is the most fun demo we're gonna do today
and also one that's been pretty useful
for me at least.
And so basically there is now an opportunity
because these bots can work as you
to go through other products and other competitors
and get a lot more insight than you ever could manually.
And then finally, shipping feedback quickly.
We all get feedback all day.
Believe me, we get more feedback than probably anyone
and a lot of it is not very nice,
so please be nice to us.
But shipping feedback quickly is super important.
And so I will go through how you can do that as a founder.
All right, without further ado,
let's meet my squad here.
We're gonna go through four bots today.
The first is the close bot.
That is the one that handles my customer.
The second is prod bot, which is what figures out
what's been shipped and helps me stay across all of it.
The third is stock bot.
That is the competition stalker.
And the fourth is protobot, which is pretty fun.
It makes some designs for me,
but it also turns feedback into PRs.
And then we have a little guest star
that I'll talk about in a second.
Okay, with that, let's get into the actual demos.
All right, here is my GrockBot interface.
I'm on a demo interface right now.
Usually there's a lot more mess on this screen,
so hopefully it's a little cleaner
and easier to parse for you guys.
The first bot I'll start with is close bot.
And there are two key things that I want you guys
to keep in mind while I go through this.
The first is remember that your bots can compound.
So you can constantly be feeding them insights
and information, and they will store it,
reference it when necessary,
and get better over time, which is really important.
The second thing is, like I said,
they can use your computer and work with your tools,
which gets really important
as you start to try to do things N10.
So close bot here,
and I'll leave a little cheat sheet on the side
for you guys, there's a few things.
It focuses on prepping calls.
So when you need to talk to a customer,
it will do a ton of deep research on them.
It'll look through our telemetry
and what we've done with them
and make sure that we have complete insight
into how they work.
And then it will finally do a walkthrough
of what they're building and their site
in case it's relevant to what we talk about.
There are a few other steps,
but I'll start by asking my bot to prep the call.
And in this case, again, I'm in a demo environment.
I don't wanna leak anyone's personal data
or any company's personal data,
but this bot will go off
and basically prep a call for this company
that we're calling Northwind.
And the way it does this,
and the way I actually do this right now
is it pulls all of the data for its usage.
So if you are a founder
and you obviously have that telemetry,
it is very good for customers
that are just trying out your product or on a free trial
to understand what they're actually doing
within your product.
And then also to understand who they are.
Those are probably the two most important things
before you get onto a customer call
because that is how you resonate with them
in the quickest manner possible.
Again, you don't have that long with these folks.
You generally have around 15 to 20 minutes
with your customers.
And so what's really important
is you get to the crux of what they need
and match it with what you're doing.
And bots help you do that really quickly.
And so right now it's going through
essentially some demo prep.
But what I'll show you is an example
of what I tend to use day to day
when I talk to founders.
This is a call prep HTML file.
Obviously it doesn't look perfect
and that doesn't matter
because I'm not sending it to anyone,
it's just for me.
But basically the goal of this
is to really understand who these people are
that I'm going to be talking to,
what their product does,
which CrockBot will research on its own
and then walking through their site.
The reason the site is really important
is actually because the site tells you a lot
about what they care about first and foremost,
which is why I like including some screenshots
of what they're building.
But more importantly,
I am constantly finding bugs on people's sites.
And it is so nice for me to just be able to go to them
and say, hey, fix this thing.
Look, your cookie banner is on top of your submit button
and that sucks for everyone.
So just go fix it.
And for them, it's really nice
because they get some feedback that's super actionable.
And so my bot has gotten very good
at figuring out what will resonate with folks.
And so I have it do this every single time.
And without fail, it has found some issue
that comes up with some site,
which is a really cool way to open a call.
So just one of those quality of life things.
As we keep scrolling,
you'll see these are placeholder numbers,
but you'll see that you get a lot of the activity
that's really relevant
to what you want to talk to them about.
So we know in this case
that these folks already have some seats,
they already have done some meetings.
This product is like a meeting tool.
They have some sort of arrangement.
And because I can look at this graph,
I can see what their usage looks like.
And so I look at this every single time
when I talk to founders to try and understand,
are they really accelerating?
And is there a specific feature
that I need to work with them on?
Or are they declining?
And there's some sort, sort of,
objection that I need to get into.
With you guys,
you will all have your own versions of this use case,
but it is so, so important to just spend five minutes
before the call figuring this stuff out.
Now, for the really cool part
that I'm a little conflicted on, recommendations.
The bot has gotten very good at figuring out
what I need to talk to founders about.
And in your case,
what you will probably need to talk to customers about.
And what that means is it will give me
a few key recommendations about what I should do
with that founder or customer on that call.
And so the most important piece for you guys
as you have your customer conversations
is to come in armed with some sort of knowledge.
And this automates getting some of that knowledge.
There are a few other pieces here.
It talks about some risks.
And I think the most important part
is that you come in prep.
You don't have to spend the hour and a half
of your time doing this, and it does it on a routine.
So every day it will look at my calendar
and look at the meetings that I have
and just prep them.
And I can trust five minutes before
I can go in and read that stuff.
Okay, so that's part one,
which is prepping the call.
Part two is what happens after the call, right?
We wanna go end to end.
We don't just wanna stop there.
And so the same bot also does a little bit
of the compounding for me.
And what that means is after every call,
it will review the granola transcript
and figure out what that person resonated with or didn't.
And you can do this with all of your customer calls
as well, right?
And arm your teams with it.
You basically have your bot look at every single transcript.
And for example, if in this case,
I was talking about some new note-taking feature
that they didn't really care about
and they expressed that within the call,
for the next one, for the preparation document,
this tool will get better at making sure
that I'm not talking about that feature again
unless it's really relevant.
And so over time, if you give this two or three weeks,
it becomes really, really powerful
because not only do you not have to go
through other people's calls
and you can leverage theirs,
but your calls just get way better
and you stop tripping over yourself
when you talk about the same things.
Okay, that's part two.
Part three is support.
Basically, we have a lot of customers
and a lot of them need support.
And right now, there are not enough people
to support all of those customers on their own,
just one-on-one.
And so you need to be able to scale up your support.
And a lot of founders leave this to the very last thing.
Believe me, I've been there.
And it makes your life very, very tough
to handle both getting new customers through the funnel
and then also supporting the preexisting ones.
And so you have a real opportunity
to automate this with Grockbot,
but the most important part
that makes this really work well
is that you spend the time empowering your bot
to actually support your customers.
And what that means is you not only connect it
to your data, but you connect it to your billing
and you connect it to the risky pieces
that you're scared to give
because that is where most of the support struggle happens.
And so that in combination with the compounding piece
leads us to this really cool fourth section,
which is activation.
And so a lot of you probably know your wow moment
at this point.
You know when your customer is activated,
you know when they're going to stick around.
And so how do you incentivize more of that?
How do you decrease time to value?
One of the easiest ways to do that
is to give people money so that they get there.
And it doesn't have to be much, right?
But just getting them across that hump
is what turns people into great customers.
And so I have hooked up this bot to actually do the work
and when an activation event happens,
where in this case again I'm using a note taking app,
if a template is shared with your teammates,
we have $1,000, which is crazy example, ready to send.
And it will automatically send that email with those credits
so that that user or that customer can leverage
those credits and they feel rewarded in that moment.
And so instead of you having to monitor
every single customer's journey and do this manually,
or taking a tool and building it from scratch,
you can just offload it to your bot
and not think about it again.
And it will run the campaign on its own.
Okay, there are a few other pieces
that are pretty obvious around the end-to-end
customer journey, right?
You want them to sign contracts,
which you can get GrockBot to go back and forth with them on.
You want them to figure out your pipeline
and actually generate it without bound,
which GrockBot is very good at.
And then you want it to actually make sure
that your calendar is built
and it can access your calendar
and do that every single time.
So these are all things that happen autonomously.
And I no longer need to worry
about 99% of this journey.
There are cases where I want to step in with a customer
and I want to understand,
hey, why didn't this call resonate?
I want to go deeper.
And for now, what that means
is I spend my time doing that,
which is actually very enjoyable.
I actually enjoy talking to customers.
So that's what I meant when I was saying
like this is a little emotionally complicated for me
in that some of this is really fun to do.
And so I have to intentionally go in
and steal some of this back from my bots
rather than letting them do all of it.
But it is a very, very cool time
to be able to scale to thousands of customers
with just your bots
instead of you having to go do this yourself.
And that way, like I said,
you can focus on making good decisions consistently,
maintaining velocity and maintaining focus.
Okay, that was bot number one.
Let's go to bot number two.
This is ProdBot.
I mentioned that example earlier
where as you start to have more and more engineers
shipping more things and unshipping more things
you're doing yourself,
you get confused around what's live and isn't.
So ProdBot is basically a bot that
does a lot of this work for you
in terms of staying across what's been shipped.
And so I'm currently asking it
to give me a daily rundown of Flylo,
which is the demo surface that we use earlier
in the PM conversation as well.
And Flylo is a tool for flights and airlines
and booking travel.
And what I like to have this bot do
in my actual day-to-day is not only go through
all of the PRs and the linear issues that come in,
but also walk through the product end-to-end itself.
Remember, your bots have their own computer
so they can use whatever surfaces you use.
And in this case, that means going through the site
that the bot is assigned to,
figuring out what has changed
and then mapping it to its experience.
So if it has a login for the specific site,
it can do a lot of this work for you.
You can see some of it popping up right now.
What you'll notice if I scroll through this
is it's giving me a rundown.
So it gives me, here's what's shipped.
These are some examples of things that have shipped recently.
It also gives me things that were unshipped,
so I avoid that embarrassment again.
And most importantly, it will give me some decisions
and some key things where, you know,
now that we ship quickly,
we don't intentionally make all of these micro decisions
all the time.
And so it tries to catch some of those
that we want to be a little more intentional about.
And so overall what you end up with
is a bunch of screenshots of the site.
So the bot is going through it.
And then also you end up with a video
of the bot basically running through the site itself.
And so that way you can see what it does
without having to do it yourself.
So if you ever want to get in deeper,
you can always ask your bot to take a video
of the surface while it's doing its work,
which has become the fastest way to verify
if something is done well or not.
The other piece is that you can connect this
to your data.
So you can actually see what the results
of different ships are
because you connected to your metrics.
And that way, instead of you having
to view that dashboard all day,
you just see it here.
And it is proactively given to you,
which I think is one of the most important parts
about bots, that they will surface information to you
as relevant and as necessary.
And so I think again, what's really important
is as long as you enable them
to have all of the right data,
they can help you make great decisions
and also help you stay across
what your team is shipping.
Okay, now for my favorite bot, Stockbot.
Right now, everyone is probably distracted
by their competitors.
And it is very easy to ship things,
but what's not easy is checking
what other people are shipping.
And so what Stockbot does is it basically
will go find your competitors,
it will sign up for autonomously
and go through their product.
And then it will give you a full rundown
of basically everything they do.
So that includes what they've shipped,
their change log, what they're posting on X,
who they're hiring for,
which is actually a really good way
to get signal on what people are building.
And finally, just who their team is
and what they're prioritizing.
And over time, this also compounds.
As they learn about more and more competitors,
your life becomes easier
because you start to get more signals
into how to position yourself in the market
and also how to counter position
to what they're shipping.
And so what I'll do right now
is kind of ask it to do a competitor pulse right now.
And basically I want it to check on notion and craft.
Those are the two hypothetical competitors we have
with this note taking app.
And so it will go off and do all of this in parallel.
And it has been doing this for a while.
So it is constantly running pulses.
You see some of its routines on the side here.
And what you'll notice is that these routines run
relatively often every few days.
But what that also means is I don't have to do this myself.
They will run the routine,
come back to me if there's something relevant
and if not it will stay silent and do its thing.
And so this is one of those ambient bots
that's really important to just have around
because it will get a lot of this work done for you
around competitor analysis
and you don't really need to think about it.
And as you get exposed to more and more
of these decisions that these companies are making,
you actually start to build the muscle around
why your decisions are different.
And you start to understand truly
what your positioning is relative to theirs.
So you'll notice it's cooking
and doing some of the research work.
What we see here is let's start with craft
which is a note taking app.
Basically it has created a little tear down in HTML
if this thing loads.
If not we can just look at it here.
It again just created a video where it's going through
and it will sign up, it'll-it'll put in its information,
it'll use a throw away email address
and then basically go through the entire flow.
So this is the onboarding flow
and it's looking through, hey,
what is different about their onboarding
versus our onboarding and it gets you to the start page.
It will also start to actually like test it itself.
So it will write in craft
which is again this note taking app
and tell me what went well and what didn't
and then give me a little overview of how it works.
Craft is not hiring actively right now.
So that's good for us to know.
And then the final thing that I've done with this one
which you know might be corporate espionage
but you know do what you want with it
and be responsible is basically I have it
email people that have churned off.
A lot of you have insights
into where your customers are churning.
So you know from Salesforce or whatever your tool is
that X customer left to Y tool.
And so you can actually go and have your bot
reach out to them and understand why that is.
You don't need to go win them back in that moment
but just getting that insight is really, really important.
And so if you have that signal, it's another good way
to course correct your positioning versus others.
You can see the same thing for Notion.
Notion has a bunch of hiring
and so I've kind of asked it to truncate
because they're hiring a ton of people
but it's still important insight to get.
And so I'm showing you these two examples
and this is one where it's signed
into its own Notion account.
It authenticated via email and is basically doing work in it
and giving me what that experience looks like.
As it starts to see the differences in what's shipped
it will focus on those specific things.
Okay, enough about this bot
but what I will say is I will share this bot after this
and I will toss up a QR code
so if you wanna steal it, you totally can.
All right, let's move on to bot number four.
So bot number four is my Proto bot
and this bot is basically really good
for just thoughts that I have in my head
when I want to get them prototyped
but also it's really good for when we get customer feedback
and we wanna just explore what this would look like to ship.
And so what Proto bot does really well
is it actually has right now
its own version of Grock bot running on its computer
and I think I've shared this a few times before
but I've given it its own account that is not it.
It has its own separate entity
and it runs through a lot of different bots,
test them out and helps me with some of my prototyping
and also QAing but I think most importantly
and the reason this matters
is because it has access to its own product
it can do a lot of this prototyping
with the context of what the product looks like today.
So if I say pull the most recent customer feedback
it will find some demo feedback, pull it in
and then it will probably just start working on fixing it
and so you as a founder can probably envision yourself
and your product piping all of your feedback channels
so that means things like X, your support inbox
or any other signals that you see in your emails
and putting them all into a single pipeline
that goes to a bot that just kicks off PRs and solves them.
That's kind of how some of our engineers work internally
which is very cool because you'll see
the full feedback loop of feedback to ship to deploy
happen in hours which is awesome
and it happens automatically and organically
which is very cool.
So I think you guys all know that adapting to feedback
is really important especially in the early stages.
You still need to make the choice
about which feedback to take versus not
because you're the visionary,
your job is supposed to be to figure that out
and in this case the fact that you can move this quickly
means the main bottleneck should be your decision making.
It should be whether this feedback should be shipped or not
and so it will basically do the work.
Right now it's saying that the customer thinks
that the share button for bot templates
isn't exciting enough and they want it to pop.
So it'll probably end up working on that
for a little bit which is very fun.
In the meantime though,
what I will show off is two final bots.
One is my YAP bot.
So this bot has one purpose and one purpose only.
It is built and exists to talk like me.
So it looks through my emails, Slack, iMessage,
any other form of communication that I've done
and it learns from how I talk.
And it doesn't just learn once.
I think this comes back to the compounding piece.
It learns over time.
So every week or so or you can set it
to whatever schedule you want,
it will take everything I've sent for the day
because it has access to all those tools
and essentially learn from them and get better over time.
So now I can trust this bot to go and talk as me.
And what you'll notice is these other bots
are actually coming in and asking it for help.
So they are proactively bringing in YAP bot
to go talk like me across all of the different surfaces.
And I don't need to do this myself.
After they do this a few times where I encourage them,
like I said, they learn to loop in the other bots
as needed.
And this bot might seem like a small one
because it's very simple, but it's so important,
especially if you're trying to do any sort of outbound
or move quickly on communications.
It's nice to have a bot you can just trust to do this stuff.
Obviously, you have some emails that are really sensitive,
so you might ask it to just keep them as drafts
rather than just sending them right away.
But even then it will learn from the delta
between the draft it created and what was actually sent
and get better and better over time,
with the end vision being that you don't need
to think about it at all.
Okay, final tip.
I have a miscellaneous bot and this bot has one job,
which is basically to be kind of a trash can
for all of the random requests I have.
Today I wanted to see if Drake's album dropped.
It is not yet, sadly, but what's really cool
about this one is I can just ask at random things
without polluting the context of the other bots.
And if it ever has any important tasks to do
where it needs to transfer that context,
I can just ask and say, hey, you learned about XYZ,
go tell Prodbot about what you've learned
so that it can get better in the future.
Sadly, Drake's album isn't out.
I checked Instagram, but who knows?
We'll get there maybe later tonight.
Okay, those were all my bots.
I think Prodbot is still going.
It's kicking off a cloud agent
and hopefully it finishes sometime soon.
But you'll notice because it has access to Grockbot,
it can actually just take a look at the interface
without needing to go find it anywhere
and it can drive it and QA it.
And as these tests happen,
it's very easily verifiable
whether it's shipped the right thing or not.
And so I think all of these bots are very powerful
and most importantly, they're all powerful in conjunction.
So when Prodbot gets some feedback,
it can pass it to Prodbot
or when Prodbot is trying to get across
some of the key product decisions that we've made,
it can pass that context off to Prodbot
so that it can sell the right stuff.
And overall, I think the key framework
that you folks need to keep in mind
is you want to start to try and productize yourself
as much as possible and give away all of your Legos
no matter how delightful they are
so that you can focus on the highest leverage thing
because that is what helps you win in a world
where everyone is shipping just as fast,
if not faster for now.
So keep all that in mind, but those are my bots.
Let's go back to the presentation.
Okay, some final pieces.
What we learned, let your bots run free.
I am going to say this forever,
give them access to as much as you are possibly comfortable
with because that is the only way
they will be able to do work end to end.
You want to invest in them
so you might find times where you need
to course correct a bot.
And with an agent, you might have the instinct of,
man, I spent too many tokens on course correcting it,
the context is polluted.
Let me just get rid of it and start a new one.
In this case, your key advantage
is that your bots get really good
as you spend more time with them.
And obviously as the team has learned internally,
the more time you spend, the better you do.
And so it's really important to really believe
in your bots the same way you would with employees.
And the other piece to keep in mind here
is the other way to invest in them
is by giving them really good skills and tool.
So it's very easy to say, hey, make this cooler.
And that could be your prompt
that you want your bot to execute on.
But obviously it's not going to do a great job
and it's going to probably get stuck somewhere
and ask you a few questions.
The more intentionality you put
behind that task in the first place
lets you repeat it over and over again
without that many hitches.
And then finally, it is worth you taking
one to two hours and just thinking about your day.
I know that sounds dumb.
I know you're like, I need to move quickly
but it is so, so important to just take the time.
I don't care when, just do it
and figure out what you want to delegate.
Because doing that intentionally
will save you so many hours
versus just spinning up bots
and hoping they do things for you
without having a true intent around what they should do.
Okay, now I have a bonus round with some tips and tricks.
I was watching the live stream earlier
and some people were like, this is expensive.
Which is true in some ways, not true in others
and we're working on it.
But there are some really easy, high leverage tips
that can make your life a lot easier.
The first is that browser use is super powerful
and it unblocks your bot from doing
a lot of things that it couldn't do
if it didn't have MCPs or APIs or some other connection.
But it's also expensive to run multiple times.
And the easiest way to solve for this
is to bias towards using APIs
and using interfaces that people have already built
or to actually kind of hack it a little bit.
You can ask your bot to look at network requests
while it does a task once
and then figure out which APIs are being hit
while it's doing that task
and then go and start to just hit those directly.
Which takes way less time, first of all
and is also way more token efficient
because you are no longer
needing to go through the browser interface
and you can just do it all directly.
So ask your bot.
And this is true for all optimizations,
by the way, that you guys have probably been thinking of
or taking notes on.
Always just ask your bot how to do it
because it will figure it out for you,
which is very nice.
But be wary of browser use more than once.
And if you can get around it and scale quickly,
you absolutely should.
Second piece, routines.
When I was looking through some of the data
around the frequency of routines that people set,
I was a little shocked
and it freaks me out how often you guys are running things.
But I think the most important part about routines
is to audit them.
And you don't have to do this yourself.
Again, get a bot to do it for you.
But the most important piece is
be very intentional about how often your routines are running.
They're really powerful,
but most people over-index on how often
they need to be running.
And they think, this routine is so important,
I need to run it every 15 minutes,
which comes out to 100 times a day.
So that is a very expensive routine.
See if you can be a little more nifty with them.
So it's not just running them fewer times.
If you really need a routine,
try and run them on web hooks
or try and run them with some sort of inbound signal
that will let your bots actually take in an input
and then run it,
rather than just having it blindly run on a schedule.
So those two things alone
will probably save you a bunch in tokens.
I know there was some demand for that.
Okay, even more final tips
before they cane me off stage.
Make a voice bot.
Obviously I love mine.
Please do this for yourself.
It is the most underrated way to get words
out to other people.
It's have a bot that copies your voice.
Import your cookies.
This lets you stay signed in
across a lot of different tools.
And it also stops you from having to constantly take over
the bots computer and help out.
Group bots by expertise.
This is kind of a personal preference thing,
but I love having my bots grouped by what they do.
And so if a bot is really good at just handling a customer,
I will give it everything customer related.
And if a bot is really good at data analysis,
that will be its sole job.
Because sometimes they're weird compounding pieces
that happen that we don't even know about.
And so when you're giving it to what feel
like completely different tasks,
you can still get some of the compounding benefits
as long as they all fall under the same expertise.
There are a few more on here, like great skills,
which we talked about, how bots can learn from you,
and then auto optimization routines.
But I think the last one is also pretty important.
Having a bot that just exists to optimize your other bots,
that means better routines,
figuring out where they went wrong
or where you had to ask multiple times for something,
and then making sure that doesn't happen again.
And overseeing how the other ones work
and improving them is way easier
than having to do a lot of this yourself.
So save yourself the time, you set it once,
and then you forget it, which is really nice,
and I don't see enough people doing that.
Okay, so that is close to the end
of what we've been doing here.
Thank you for your time and attention.
I am gonna leave a QR code for stock bot
if you guys wanna steal it, it's a fun one.
You should try and use it on us.
I think that would be cool.
I wanna see what people come up with.
I really appreciate everyone being around.
I think one of the coolest parts
about these bots and where they're headed.

is just how creative you can be.
The amount of different bots and different combinations
of bots that I've seen has really blown me away.
And I think founders are the perfect group
to figure out the edges of this entire thing.
I don't think we've hit the hit the line
and as we get closer and closer to figuring out
what these bots can actually do and what they can't,
we're gonna be looking for signal from you guys.
So really appreciate it.
If you are on the live stream,
we're gonna send you back to Matt Lauren and Roshan
to go and continue cooking.
So appreciate it, thanks.
Okay, for those of you here,
we will take 15 minutes and answer some questions you have.
So we'll have some facilitators who maybe have mics,
if not just yell really loudly and we'll go from there.
So raise your hand if you have a question
and we'll get after it.
Yes, up here.
The idea of treating bots like humans.
So you said you should spend one to two hours
setting bots up.
Do you have a framework or suggestions
as to how you actually do this?
What's the way in which you create jobs
or roles responsibilities for bots?
Yeah, the question is like,
what's the best way to set them up essentially, right?
I think when you're trying to be intentional,
there are two things that matter, right?
One is figuring out what you need them for.
So like actually inventorying every single thing
that you need to do is pretty important
because then you can start to group those things.
And like I said, I like my bots to be experts
and so they will have a very specific domain
that they're good at and then they will expand their scope
as I have more questions for them outside of that domain.
And so I trust them to do things
outside of what they've been assigned.
But for the most part, I try to group them in,
your finance bot or your job is dealing
with customers or your job is like marketing generally.
And that's worked for me.
Some people really like having the chief of staff bot
that manages the others.
I personally am not a fan,
but it really comes down to personal preference, right?
If you're someone who just wants all of the work abstracted away
and you just wanna talk to one, it's a good way to go.
But if you really like being a little control freak
and getting into the weeds,
I think having multiple bots that are all experts
that you can then train up is kind of a good way to go.
And I would love to give them little XP bars
or something as you go, but we'll see.
Yes, right here.
I just had a question about managing
broad bots on multiple computers.
Is there any tips for that to work
kind of and not get confused in the whole ecosystem?
Yeah, yeah, yeah.
So I think like a week ago,
bots would get pretty confused
across multiple computers if you had them.
That is something that we're actively working on,
making better.
I know there are a lot of folks
that used to be on other setups
and had stacks of Mac monies lined up
and they wanna utilize all of them.
So it's something we're actively investing in.
Right now, that should actually work a lot better.
So if it doesn't, we should chat.
But yeah, thank you for the question.
Great.
Hi, when it comes to serious enterprise use,
and you have actions and decisions,
so is there a way of putting those,
having those decisions follow a certain
deterministic rule or policy
that is outside the model?
Like to ensure everything is safe.
It doesn't do anything without going through that rule.
Yeah, so the question is like,
how do we make sure that models act
deterministically in this tool?
Well, they inherently don't, sadly.
But the easiest work around there
is probably to make sure that your model's just right code
and code itself is inherently deterministic.
So as long as you kick off cloud agents to write in code
and you give it a little decision tree,
it can basically play the deterministic part
and then you just tell the model to call that function
every single time you need it.
You can tell bots to ask other bots for permission
if you need to, but I think the better way to go about it
is just to give it a rule and say every single time
you need to make a decision related to this,
look at this flow chart and then execute it.
And as long as it has some sort of code to execute on
that is verifiable, it will be able to
cosplay being deterministic.
Yeah, cool.
Yeah, yeah.
Two answers to that.
The question was, is there like a good way to import
all of your setup from other tools?
One, working on it right now,
I have pretty cool bot that I wanna share
in the next few days that should make this a lot easier.
But two, I think part of this is making sure
that all of your connections are interoperable, right?
Like there's, you have your collection of MCPs
and APIs that you already access
and right now they're probably dispersed
among a bunch of different tools.
And I think the easiest way to bring them all together
is the same way you're kind of doing it
with agents and skills, right?
Like where you either have them SimLink
or have a direct source of truth to handle all of them.
On the Linux piece, it makes it a little more difficult,
but I think there's some cool integrations you can use.
Like one password is an easy one
where if you just import everything in
and then use the MCP, it does a really good job
of figuring all that out for you.
And then the other is cookies like I talked about
where you can import all your cookies from Chrome
or your browser and then you just use those sessions
within your box.
Yeah.
Yeah, yeah.
So there are a few authentication pieces
that we're still figuring out, right?
And then the other pieces,
we have some companies who are like,
hey, we don't actually want bots on our platform
or our goal is to be good ecosystem players overall.
And so those are challenges that are going to exist
across any tool in this ecosystem,
but it's something we're actively working on.
So it's a good question.
Yeah, it's when we're asking ourselves.
Hello there.
Yeah.
But then I ran into problems with the CPUs,
like with keys, all the, do I have to do it?
Why can't you just drop the MCP, go in the browser,
and just figure this out and just give me what I want?
Right, so is it happening or like,
what am I missing or like, how can I do it?
I think, so the question is like,
hey, this kind of isn't working for me, right?
Like I'm trying to authenticate into this tool
and it's asking me for all this stuff,
like how do I deal with it?
I think it just depends on your setup, right?
And I think one of the best ways to use Grockbot
is to just ask it to figure this stuff out
and not be super prescriptive on the tool.
So for me, the reason I landed on Granola in the end
was not because I'm a huge Granola fan,
it's literally because it was the easiest form factor
for Grockbot to pull transcripts
and get me the insights I needed,
especially for things that are happening on device.
Over time that might change.
And my bot that is working on optimizing my other bots
will probably figure out like,
hey, wait, this is now a first class,
like first party integration,
and I wanna use that instead.
So I think being less prescriptive on the tool should help,
but also your point is heard, right?
Like it's, it is really frustrating
when you're trying to connect to a tool
and it just doesn't work on its own box.
So I think parts of it are just getting more creative
right now as we're in the early stages
and the cope is eventually we're able to handle
all of this demand across all these different tools
as we start to introduce more plugins in the marketplace.
Great question.
You, yes.
Not yet.
Yeah.
We'll see.
Yes, right here.
Yeah, hey, so a lot of legacy services
either don't have APIs or have API access
that costs much more than the UI access.
How fast is Grockbot going to get
where kind of the answer that we're looking for
instead of using APIs,
it's just as fast as kind of going through the UI flow.
Yeah, so the question is like,
how do we make sure that going through the UI flow
can eventually get just as quick as using an API?
Yeah, or are there things that we can do to make it
just as fast?
Cool, I think there, that's a good question.
I think there are two things that help optimize here.
One is trying to go headless as much as possible.
So using a headless browser to just do a lot
of this work quickly for you,
and that way you don't have to navigate UI.
You just basically give it commands to click pieces
in the DOM instead of having to look at it, screenshot,
make a decision, rinse and repeat.
And the second thing is we're doing a lot of work
on our model to make them faster
and better with computer use.
So I think overall the promise of being faster
than an API is a little tougher,
especially as you add more layers to it
because inherently the performance is worse.
But our goal is to get as close as we can.
And over time that's kind of where we're headed.
Cool, yes, right here.
So the second question about like helping out,
how do you make sure like,
in order to communicate with each other
and then also they choose like the right?
Yeah, so the question is like how do we optimize
for multiple GROC bots?
And then if we're trying to be really aggressive with them,
how do we make sure they make good decisions
on themselves, is that right?
And also token optimization.
Okay, so the first thing is right now,
we have this feature that you all can use
that is like putting bots in a group chat.
And that comes with pros and cons.
The pros are that you get a bunch of bots
that are really eager to do a lot of work for you
and they can work together,
which is cool for very complicated tasks
where you loop them all in.
The con is that they will all talk.
They all love to talk.
And because they're all proactive and ambitious,
when you put them in a group,
they will start to speak over each other
and they will work really hard,
but they will also get really expensive
as they start to do more and more.
And so I think the part to be cautious about
is whether that's the right form factor, right?
Sometimes you just want to bot to tag
two other bots once and work with it separately
versus having them all in a group chat
and watching them all respond like crazy.
So I think that's one that you probably want to stay away
from for now.
Two is on the token optimization and model optimization.
I think we're obviously constantly working on
making sure the bots are using the right model
for the job across everything like browser use
and even internal bots.
But I think the harder question is
what happens when you're shipping, right?
And which models you use?
The way that gets solved today
is just by using cursor or grok build
and using those cloud agents.
You can specify which models they use yourself
if you want to tune it.
And you can even give them defaults
if you just ask the model like,
hey, I want you to use XYZ model
for this specific task every single time.
The final piece I'll say is not many people know this.
You can just message your bot and tell it
to forget things.
So if you're finding out that it's, you know,
referencing a piece of context way too much,
you can just say, hey,
forget all things about how we generated
your profile picture.
Just don't ever think about that again.
And it does a pretty good job
about clearing that out and becoming better,
which also helps with that token efficiency.
Yeah, thanks, man.
Let's go back there, white shirt.
Yeah, the question is how would I rate or rank?
Okay, how would I rate the bots on the marketplace?
Yeah, so how do I evaluate
whether a bot is valuable on the marketplace?
So we have our internal marketplace,
x.ai slash bot slash marketplace, I think.
And that marketplace is really cool and useful
because it is all hand audited by us right now.
And so we have a lot more bot templates
than you can imagine,
but we're making our best effort to go through
and find the actual most valuable ones.
And then if we get consent,
we try and optimize them for folks.
And then if they're good enough,
we put them on our marketplace.
The easiest way to check whether a bot is doing a good job
is honestly to just try it yourself for the most part.
And most of these bots,
you can ask it an initial setup question,
say, hey, what do you do and how do you do it?
Without actually having it run through the whole process,
and that way you can do a little bit
of that self-selection early.
But the one thing I can't say
is at least for our marketplace,
you can probably trust the bots in there.
So they've gone through a pretty thorough review,
which not only includes human review,
but also we have bots that review the bots
as you guys saw, which is pretty fun.
Cool, yes, right here.
Got it.
So the question is how do bots talk to each other
across people, so across people and across accounts.
It's something we're working on
and we're trying to see what the best form factor is
across all of it.
Does not exist in our capabilities today,
but it is a very cool use case
and we are excited for that one.
Yeah, back there.
First of all, he said codex, so everyone get him.
He's kidding.
I think, good answer.
Okay, so the question is how do we use local
and the bots computer at the same time?
The first thing I'd say is go to your settings
and turn on local execution.
So you can just allow it by default if you want.
That's kind of the easiest way to go.
And that lets you do some cool stuff.
It lets you send messages from your computer directly.
It lets you drive your browser if you really want to.
And so you can do local execution.
The reason we bias towards moving so much of this
to the bots computer is because we think that's the future.
We think as models get better and faster
and obviously the computer is getting stronger,
you get a lot of your work done
and delegated off of your device.
And there are a few reasons why doing things locally
is kind of janky even if it's convenient
in their near term.
And those are one, especially for certain apps
unless they've been solved for,
they will just pop up on your screen while you're working,
which means you can't do things in parallel.
People are trying to fix that,
but it's not a perfect solution.
And then two, they use up your computer's resources,
which gets really annoying
if you're doing some high intensity stuff.
But it's definitely heard and you can use it right now
A bots computer is actually one VM
with a bunch of different instances for each bot.
The same way you would have different desktops
on your computer and they share files.
So you can have them read each other's files
and they will actually proactively do that
if they need to for a task.
But their context window is obviously not the exact same.
That's what you're asking.
Cool, thanks.
Yes, right here, got it.
So the question is around the interplay of cursor,
cloud agents and then grok bot, right?
We actually have a first class integration
which is kind of a perk of having grok and cursor be together
where you can kick off cloud agents
and grok bot can transfer only the relevant context to them.
And that means that they can work independently
as a cloud agent and then come back with the results.
And then what grok bot can do
as long as it has access to the environment
is it can actually go and QA it, it can test it.
It can even look at the PR or whatever has been shipped
and work with it directly.
So I think the most important part
is actually figuring out when to use a cloud agent
versus when to just use grok bot.
And I think anytime you have some sort of complicated task
or something that you want to ship
and you want full control over which model's being used
and you want to be able to go in and really change things
that's when you use a cursor cloud agent.
But I think trying it out a few times
will actually get you the intuition you need
to answer that question for yourself.
So highly recommend just like spinning up a few
and getting them to ship things.
And you'll find that it does a surprisingly good job
with the orchestration piece
when it hands things off to cloud agents.
You just have to specify once.
Cool.
All right, that seems like all the questions we had.
Few more really quick notes.
One, star base.
You can win a trip to star base, which is very cool.
I don't know if I'm eligible for this or not,
but I hope I am.
This is a pretty simple challenge
and it is a very cool like once in a lifetime opportunity.
So highly recommend trying it out.
I don't know how we got this approved,
but basically we have a challenge post
that went out already.
You also get a plus one.
Okay.
Yeah, can you hear me?
Okay, awesome.
I can't hear them.
I can't hear their sound yet.
Hopefully.
Okay.
I can hear you.
I just can't.
Yeah.
Oh, I hear sound now.
Okay.
You got us Jenny?
Yeah.
Okay, cool.
Awesome.
I got you.
We just gave an intro,
but we'll do one more quick intro.
I think just to catch everybody up
and then we'll jump into the conversation
if that sounds all right.
Sounds awesome.
Awesome.
Cool.
Jenny, thanks so much for joining us.
You lead an incredible community of women building
and exploring with AI.
You have a background spanning venture capital technology
and community building.
And from what we see have developed
sort of an incredibly thoughtful approach
on building a media business
across platforms, channels and formats.
We're obviously building a business today.
And oh, I see we're joined by a special guest too.
So thanks for bringing on helping grow our audience.
Our first Grock baby.
Oh, I love it.
I love it.
Gotta teach them young.
Well, we're excited for you to join us.
I think we might have some sound issues in the stream,
but we'll see if we can fix those.
Okay.
We're building a business today.
I don't know how much of our stream
you've caught over the course of the day,
but we're working on building a business
powered by GrockBot.
We know you're a really active GrockBot user.
And so maybe we'll just love if you could start off
by telling us a little bit about yourself
and how GrockBot has been useful to you.
Yeah.
So as everybody can see,
in addition to being a VC,
being a full-time creator,
building an entire media empire,
I also have a brand new baby.
And so I have been able to basically double down
on building all of my companies because I have GrockBot.
So most recently,
I just built out my brand new media agency.
So I have an agency called Gemma Media
where we basically build executive brands.
And I've got 22 GrockBot employees.
Oh my God.
So it's such a really interesting time
because you normally wouldn't be able
to do this kind of thing.
And now because I've got all of these GrockBots running,
I'm actually able to build businesses,
well, raising a baby at home
and doing things all at once.
So yeah, it's been pretty incredible.
Tell us a little bit about these bots.
22 of them, I'm really curious.
Yeah.
So I have seven different businesses that I run,
ranging from Real Estate to VC Fund,
to my media business and my agency.
So we've got,
and Shubb did not like the Chief of Staff model,
but we do have a Chief of Staff.
We have Shukong Master Chief, my husband named that.
And so Master Chief basically is the Chief of Staff
that oversees all of our different businesses,
ranging from overseeing our tenants and guests
in our rental properties,
all the way to managing our agencies,
all the way to even managing inbound brand partnerships,
things like that.
And so it's basically one Chief of Staff
that is running everything Jenny Co.
You've set up, it sounds like a bunch of bots.
Are these sort of specific to specific parts of the business?
Like, yeah, why so many and what are they all working on?
Yeah, so basically we have them all doing
kind of an individual task.
So we've got, for instance, Boxie
is just the inbox monitor.
We've got Scribe, Scribe is in charge of,
every time we whisper flow,
every single meeting that we do,
and then that note taker intakes it,
and that's in charge of taking every single thing
from every meeting and delegating it out to subagents.
We've also got like an entire CFO setup on there
where it's doing all the different bookkeeping finances,
keeping track of my receipts,
like when I'm out, for instance,
they basically all kind of do a core function.
And then we use the Chief of Staff to spin them up
and do essentially that group conversation
that Shubh was just talking about.
Like they literally jump in on a project altogether
and essentially chat and get everything spun up.
So I lost some ideas.
We've been pivoting a little bit as we kind of figure out.
So today we're gonna build a business in 72 hours
and obviously you have some experience there,
both building and advising.
And so we're really keen to just learn from you a bit
and think about how we should improve our idea
or kind of change our idea.
So I think we're working through a couple of things
on our side around in-person pop-up
and thinking about how we might allow folks
to connect to local businesses.
That's been a really interesting angle
as part of the day.
Also exploring some ideas around how we might let folks
who aren't located near us maybe jump in and join the fun
perhaps virtually in some way too.
And so we'll just love to learn
from your experience around events, businesses
and like planning these types of things.
What comes to mind?
And we're like, we think we wanna go software for pop-ups.
Great idea, not great idea.
Things that might work, things that we should keep in mind.
Yeah, so I saw you were just on with Eric, my buddy.
And he had a lot of great ideas.
So the last thing I saw that we were like walking
through the inventory of merchandise.
And so I think what would be really cool to understand
is like right now, where are we at
with like the actual pop-ups and the scheduling
and everything that we actually wanna execute?
Like can we get up to speed on where you left it off with him?
Yeah, yeah, we're like very behind.
So we have a lot of ideas.
Yeah, yeah.
We have a list of like places we can reach out to.
We haven't actually done any of that work.
So we could start to think about that,
but that's where we're at.
Okay, so do you guys mind just turning it up?
Can the producers turn up the volume a little bit?
You guys, your mic is still down a little low.
Cool, can you hear me all right now?
Yeah, now I can.
Awesome, I'll speak up a little bit.
Basically, we still have to reach out to places.
We have some good venues that we could check out,
but we don't really have the logistics
of how we work this thing out.
Okay, so the good news for y'all is that
in addition to building a VC fund,
I also built one of the biggest pitch competitions
in the world.
So I'm very familiar with venue selection.
So in terms of the venues that we're looking at right now,
we're focusing on SF, right?
That's right, yep.
Yep.
Okay, so have we done any sort of like putting together
a criteria with one of the bots
of identifying what we need in the venue?
Yeah, so loosely, like, hey, we want about 100 guests.
We want kind of a warehouse space.
We want it to be nice and open
so we can have like sort of a gallery
inside that, in that space.
Maybe a little bit more than a hundred I think,
but that's about the extent of that.
Okay, so the first thing I can tell you
as an event producer is like,
we need to think through venue logistics to add to it.
So before we have a botgo search out for venues,
we need to get a little bit more detail.
So number one, what's the timing we want to do?
I'm thinking sometime in the next month.
Maybe next month, maybe a little bit longer.
Maybe a little bit longer.
I have to get some permits or something on around.
Yeah.
Well, so that's the next question I'm gonna say.
So as somebody who's produced a lot of big conferences,
permits are a big thing in the state of California.
They love permits.
My last conference had more than a dozen
different types of permits.
So are we thinking alcohol is involved?
We'll say no, just to make things really simple.
What if we said no?
Okay.
You do streamline things quite a lot.
Do we want food?
Okay, yes. 100%.
Okay. We want served hot food or are we doing grab and go?
We're thinking served hot food, yeah.
Okay. So these are all the details
that we're gonna wanna put into a bot
to do our venue search, right?
So venue searching used to take event producers
like hours and hours and hours to scan the internet
and put in all this different criteria.
But the best thing we can do is build basically
like I call mine scout.
Like this is like the search bot that can go out
and identify which of the venues
in the County of San Francisco meet this.
So what I would do is like sketch out our capacity
which we've got.
Do you want this to be in the evening?
I don't think we've decided.
We have no decided.
Okay.
So let's do, we want the timing
because not all venues, some venues are on nine to five.
Some venues are closed in the evening, right?
And then we're gonna wanna put like
what types of food we're gonna want available?
Do we want kitchen facilities on site?
So I would say yes.
Like we were thinking originally that
some of the prep would be able to be done on site
but I can see how that would add another logistical
later versus if we like hired a catering company
that then brought the food in, right?
Okay. So if we're doing a catering company
then you need to make sure that our venues
are gonna accept our catering companies.
So a lot of venues in the city of San Francisco
are gonna have preferred vendors for their catering.
So are we trying to do like bespoke
or do we wanna just kinda lock into
what someone's already working with?
Ultimately, we probably care more about having
a really great space and then kind of working around
the constraints that the space might give us.
I don't know if that's a great way to think about it
but I think, you know, we wanna create
like a really fun environment
and I think have that be the leading thing here.
And then, you know, if there are things like preferences
that we need to consider, maybe that's like something
we could ask at Gronk Bot to help us go research
and figure out is that kind of where you're going with that?
Yeah, I think that makes the most sense
because the way, like the more specific
we can get in the venue search,
the faster we can actually get to a result
of like the availability, right?
So the city of San Francisco,
beautiful place 49 square miles.
It's a very hot spot to find a venue
and there's also a lot of legality
when it comes to the city of San Francisco.
So I think putting together a bot that does like
our venue scouting specifically
would be like the next step here.
And that bot can work in conjunction
with the other details of the event.
So for instance, do we have an event planner bot yet?

Ooh, what would we do with an event planner?
Well, so how many pop-ups do we want to do?
Just the one?
Yeah, do you want to start with the one
and then kind of figure out what we do from there?
OK, so then what I would want to do
is it's almost like the chief of staff model, right?
So we basically reverse engineer an event producer.
So we have an event planner that
works over top of what would be their coordinators
and staff underneath, but then we
build this as a bot model.
Yeah, I like that.
So it seems like we should maybe start
thinking about a team of bots that could help us go and do
all this work.
We have sort of like an event planner with context in the event
and then maybe split out some bots that are more task.
I'm curious, is it email?
Is it phone?
Should we think about giving our bots the ability to call
places?
How would you think about the actual bot interactions
with these venues and tying that into the AI
and what we've done with ProcBot so far?
Yeah, so I would definitely integrate it by email.
So like for instance, I've got a bot that runs reselling
all of my clothes across like Poshmark, Depop,
Mercury right now, interfaces.
One of my bots right now is reselling clothes as we speak
and negotiating bids with potential buyers for me.
So I set a framework around how to negotiate.
So we can not only have a bot that goes out
and ask for RFPs for venues and availability,
but we can even prompt it further
to be able to negotiate and go back and forth
because we'll give it parameters around what our budget is.
So that's one of the best things that I've been able to do
with Mycroft Bot is it knows how to negotiate.
Totally.
OK, and how did you teach your bot to be an A-plus negotiator
and what tricks can we steal so that we have a really good
luck with the negotiation partner?
Well, I think that we need to first off establish our budget.
Have we budgeted out the actual event?
We have not.
And so I think what we need to do
is we probably want to establish a baseline first, right?
Like establish a baseline for what venue rates look like,
what maybe food rates look like, and get a sense
of how all this things can tie together.
And so maybe we kicked off some market research
on ideas earlier today.
But maybe you also should kick off some market research
on the price lines.
Or maybe you have some knowledge you can share with us.
Yeah, so I've done this quite a while.
SF is one of the pricier cities for this type of thing.
So I think that an event planner bot would be great for this.
So you can actually ask the event planner bot
to create the first production budget, right?
So we want to have the F&B costs in here.
We want to have the marketing costs.
We want to have any actual production costs for staffing.
Even the GROC bots can get us like most of the way.
You are still going to need human beings doing
the registration and the security.
So I think that starting with the event planner
is actually making a production budget as a template
and saying you are a senior event planner
in the city of San Francisco creating a budget
for an event that meets blah, blah, blah, blah, blah.
And then with those details, it can then go out
and request RFPs to the venues
that actually meet those qualifications.
Okay, I'm kind of going into a GROC bot right now.
You are a senior event planner
in the city of San Francisco creating a budget
for a pop-up that will have 100 to maybe 200 guests
I'm going to make that up.
We can all change it later.
Let's say pop-up event.
You want to specify event?
Okay, pop-up event.
Because a pop-up could be outside too.
And we need help with a budget that includes venue costs.
So it includes venue.
Includes food and beverage.
Yep, food and beverage.
We also want to make sure.
We also want to make sure.
We're just getting a screen share up here real quick
so everyone can see the conflict.
Roshan's getting up.
I worked to make it a little bit bigger
once we get the screen share on.
Hopefully I'm hearing we can get that up in a second.
So we'll let the team take care of this
and we can keep working through this problem.
Yeah, okay.
So we need help with budget.
It includes budget that includes venue costs.
And sorry, go ahead.
So let's go venue, F&B, so food and beverage, right?
Yep.
We're also going to want to know staffing costs as well.
So some venues are going to provide staff with it.
Some venues don't.
Depends on where you're working.
We're also going to want to find
are we baking the marketing budget into the event
or is the marketing budget a separate line item
we're using for the entire event?
We'll consider that a separate line item for now.
We just want to focus on the logistics
of actually throwing the event
and then we can figure out marketing separately.
Okay.
Okay, so event promotion typically would go
in the event budget too.
So if you're doing any sort of like invitation design,
meta ads to promote for signups or anything like that,
you can either house that in your marketing budget
for the company or in the event budget
depending on how you structure the business.
So both of them work.
It just depends on what you see it.
Do you see it serving only the event
or do you see it serving the broader startup?
Cool.
Okay, I'm going to just say
let's ignore the marketing budget for now.
And hopefully that helps us simplify the problem
a little bit and kind of get to a baseline maybe
and we can always add that in later.
Yep, sounds good.
And then some other basic costs like do we want music?
Do you want a DJ?
Do you want AB?
Okay.
Yeah, my understanding is a lot of venues
will have different options there too.
So maybe we need to like let's plan for AB needs
as part of this, but we don't need to be too complex.
Yeah, so I think we need to entertain.
No, but I do think you want the ability
to like have a microphone so that
because we want to we want to speak to the audience, right?
Yeah.
So part of this is teaching them about this launch.
So you're going to want at least some basic AB.
So like a microphone and speaker in the very least
maybe we don't need a whole stage
but you'll want something with basic AB baked in.
Cool.
Okay, so I have I literally I've just said
we want mics to be able to speak to the audience
and maybe we'll just have the box
figure out what to do with that.
Cool.
Yeah, because it'll say often
once it starts negotiating with RFPs
it'll see who has AB.
Because the goal here is the most vendors
we can get in one venue
because otherwise we're going to have to create a bot
that goes and scouts vendors separately.
Okay, awesome.
I'm going to fire off the spot
and maybe we'll see what it comes back to us
in terms of budgets.
I know we don't have a ton of time with you today, Jenny
but I'm just curious what we just did was really helpful
for like one portion of how we might want to run this pop up.
Right?
Like let's go get a bunch of research on venues
for example, could you give us like one of the two
or three other big things that we should maybe create
equivalent box for and we can go up with some
of those prompts separately.
But just curious.
So we have our venue budget bot.
If we're like creating two more bots to make sure
we really don't mess anything up.
What would you recommend?
To make sure we really don't mess anything up.
This is, I might be the first person to ever say this.
This might be the first Grockbot, a permit Grockbot.
So I would have, so I would say, can you hear me now?
Yes.
Okay.
So I might be the first person to ever say these words
a permit Grockbot like red tape Grockbot.
We need a Grockbot to go research
and see if there's anything legally needed
for the execution of this event.
Because SF, they keep, they run a tight ship.
We don't want to skirt the line.
So I think a permit Grockbot would be great.
Yep.
So, and this is a great tip for anybody
that's doing any sort of event.
Like regulatory research like this
is such a pain in the butt for anyone
that's ever doing an event,
especially if you're doing them in different cities
because every city is going to be different.
So having a Grockbot that can just check policies
for anywhere where you're executing on something
is a great idea.
You can also obviously build like a Grockbot
that's doing other things internally like this
that would check like employee policies
or any sort of policy related work.
I think a bot would be a great source of truth for that.
Yeah, which would definitely be like a policy bot.
Maybe that helps us navigate a lot of that
and can maybe even help us review some contracts.
Well, I don't know,
I don't want to get like an actual advice there,
but yeah, at least your first pass
so we know what we're reading.
That could be helpful.
Yeah, yeah, cool.
Yeah, I would say,
well, I was gonna say in terms of the,
I've got a lot of friends that are, you know, attorneys
and they say that you should treat the AI
as kind of like a second year law student level of review.
So it's somebody that knows what they're doing
but they don't necessarily know the nuance
of how things operate.
So it can read the legal ease of what the law says,
but like what's the going rate in the city
and in the industry of how things shake up, right?
So I think it's great for a first pass at contracts
but we would want somebody with experience to overview
beyond that.
Cool, awesome.
I think that makes a lot of sense.
I think that actually gives us a lot to work with
for the rest of the evening
and we're getting close to the end of the stream,
but it is actually really great having you on
and talking about some of these things that, you know,
honestly, you have to be kind of an event planner
or have gone through these experiences
like you have, right?
To know about,
and maybe that's the power of being a human being
and like having the experiences
is you know how to direct the AI
and what you get for it.
Yeah, so the thing that I tell people,
the reason why I've been able to build out this agency
on such a lean, you know, budget with all these AI
is like do it yourself as humans,
build it out with what humans would do
and then kind of reverse engineer.
So I think for what you guys are getting towards
is a great direction.
I'm curious then,
so how are you, how are you thinking about this
as we wrap up the day like going into day two?
Like what do we want to execute on?
Yeah, I'm actually,
we were actually just thinking about that right off camera
a minute ago.
What we want to do to kind of close off the stream
is figure out like what long running agents
that we should try, you know, just like executing on tonight.
So I think this budget agent is definitely one
where maybe we can exchange an email
to make some progress.
Any other sort of long running agents
you think we should try getting up tonight
and like working while we sleep
as maybe some parting thoughts?
Yeah, so I definitely would say like
the Scout venue searching.
I think like we should definitely make sure
that it is searching, identifying and drafting emails
for in the morning we can review everyone
it wants to outbound to.
So that's like my favorite thing.
Like I constantly just have like the agents
drafting emails for me that are waiting in the morning
and I just have to hit send.
I also would say like we also need to start kind of like
figuring out how we're gonna get people there.
So an easy thing to do is like
I give my agents access to Metta
and like they log directly into my Instagram.
So like let's start sourcing an invite list.
Like that's the most like basic thing is go through
all the people that are verified that follow you.
You can even source like I'll tell my agent
find the 50 accounts that follow me
with the highest amount of followers that are in this city.
Let's start making that list.
And then we can pre-draft to them outbound
so that we can get the most important people
to be coming to the event
and also sharing about the event.
I love that.
We're fortunate to have a really large audience on X today too.
And so that might be a great source for us too
to go and poke after.
But I guess with that Jenny, thank you so much.
This is incredible.
Thank you for being such a great GrockBot user too.
I saw a screenshot of your boss earlier
was trying to get the screen share working
and I wasn't able to get it working.
But we just make sure we post that somewhere.
It is a masterpiece in terms of how you set up
the sidebar and have all these different agents
kind of running your businesses for you.
Absolutely.
I'm super excited to come to this.
You guys are gonna have to do this.
You're gonna need to make the merch.
This is happening, right?
I think it's happening.
I mean, I feel like we're, yeah.
Hopefully we can figure out all the logistics
in the 48 hours we have left.
But yeah, we're gonna have to figure this out.
There you go.
Thanks so much.
You're doing awesome.
Thank you.
Awesome.
Thank you again.
I'll see you guys.
Awesome.
See ya.
Cool.
All right, well, I think we have Jenny,
honestly, super valuable, super great.
I think charismatic.
And I think we have about like 10 minutes left in the day
when we wrap it at 5.30.
So kind of update where we're at.
Really spend the first day, I think,
kind of fleshing out these ideas.
We aligned, shifted a few times.
Yeah.
We might still be shifting.
We brought the whiteboard in.
There's some things written on the whiteboard,
but we still have a long ways to go.
And maybe before we close out,
we could talk a little bit about what we learned
from Jenny or where we think we're gonna take things
in the next day or so.
Kind of just reflecting on the day.
And maybe we could all end with a few
short thoughts on the day.
My immediate thoughts are,
one is I think idea generation is really hard.
I think we've definitely spent a lot of time today
poking around different ideas.
And I think the other thing is
there's a lot of wisdom in the crowd.
Like everyone we've brought on
has had some unique perspective.
And so kind of what I'm excited to think about,
while our agents are running overnight,
maybe we should do some of this too.
Is like, how to really lock in
on sort of one or two things
that we want to really ensure we can build.
In the next little while to make this all successful.
And so yeah, I think figuring out how we get
to some of that clarity will be really great.
I'm just, I think the,
we've done a lot of like broad idea brainstorming.
And so I think kind of like,
how do we focus that into something
that we think could be really fun for people?
What sort of, what's up with mine?
Totally agree.
I mean, it was cool to see like,
we got our Slack spun up,
we got everything connected.
Got boss right into Notion,
got GitHub set up, got Vercel set up.
Started with a dashboard
started with like a very V1 lander
kind of iterated on that a little bit
started moving forward.
So I feel like we have a lot of like set up stuff done,
but we definitely need to kind of lock in.
So I feel like we're gonna have to have to do that.
Maybe start tonight.
Honestly, I don't know.
What do you, how do you feel?
I think so.
I feel like after talking to Jenny,
I feel like it's quite eye-opening
how difficult it is to do a start pop up in SF.
Yeah.
And I wonder, I guess I'm trying to think about like,
is this the right problem to go after?
Yeah.
So yeah, a lot of thinking I think we need to do,
but yeah, hopefully we'll come back tomorrow refreshed
and we'll lock in tonight and think about
what we're gonna build.
But yeah, I think we learned a lot from our guests.
I know.
So yeah, cool.
Awesome.
Yeah, I mean, it was a busy day.
And again, we're in San Francisco
or Gold to Build a Business today.
And then the next two days as a part of Dreamforce
at Grockwok Galaxy, just a wrap before we close it out.
The Galaxy Live Build Challenge,
the Grockwok Galaxy Challenge,
corresponding to the live stream.
If you create a template,
you submit it, go to atgrock.
We, it looks like we might have lost video.
And you can submit your template there,
but thanks for joining us.
And I think we're out.
See you tomorrow.

