Hiring Employable Agents + Startup Battlefield Finals — Howie Liu, HyperAgent

Please join me in welcoming the chief executive officer at HyperAgent, Howie Liu.

All right. Hello. Um, who's excited about the end of this event? This, this has been an amazing few days, amazing talks, amazing, uh, Gary talk right there. And, uh, by the way, I've never been introduced with so much bass. I am very, very pumped in this moment. Um, so I consider myself a product thinker at heart, meaning, you know, as much as I like to spend time on the technical side of building and, you know, I, I love to think about architecture and algorithms and so on, you know, really, I'm a product form factor thinker at heart.

And, you know, what I wanna end today off with is a brief and kind of high-level overview of how I see the world going with agents as it relates to product form factors. So I'm calling this hiring employable agents, and it's really a reflection on as these models have gotten better and better, and we've heard from some really, really great speakers about, you know, the latest, uh, frontier models, whether GLM 5.2, uh, or others, and how to coax the most, um, performance out of them. And really, this is not about the technical deep

dive, but how do we think about these from a product-building standpoint? So who am I? I'm the founder and CEO of Airtable, and also now HyperAgent, which is a product within Airtable. And as many of you already know, Airtable is basically a almost laughably horizontal platform. Uh, when we in-in fact set out to build the company almost twelve years ago, we got, you know, basically this negative feedback from virtually every investor we talked to early on, which was, you know, "You can't possibly go out with a horizontal platform approach and try to build something that is

everything for everyone, like any type of app that people wanna build. You know, you can't solve all of those use cases." And, you know, I think we ended up doing a pretty good job of solving for many of them at least and solving for a large builder community, and we're basically applying a very, very similar philosophy to agents now, and I'm gonna break down how we're doing it. So, um, when I zoom out and think about the progression of models and therefore the product form factors they enable, I really see it as this spectrum, right? So we

went from basically completions, which is, you know, you give a model, one of these LLMs, a, you know, string of tokens, and it just, you know, continues emitting more tokens until it gets to a stop token. And more or less, it would just kinda complete your thought, right? Or complete, um, the pattern. And, you know, un-until InstructGPT happened and enabled a chat form factor, it was, like, arguably powerful but more narrowly applicable. Once we got to a higher level of model intelligence and then also the kind of innovation of

InstructGPT and kind of tuning these models to actually have more of a conversational back-and-forth behavior with users, then we enabled the chatbot, right? And of course, we all know ChatGPT was an awesome breakout product. We're now all very, very intimately familiar with that form factor of AI. The next level I call an agent, right? And, you know, I think agent has been very overloaded. Um, we've probably talked about that a ton over the past few days. But my definition of agent really is the same as the Anthropic definition, which is distinct from workflows. And an agent

is really a model that recurses upon itself, right? And has, you know, the, the open-ended set of tools that it could call, of reasoning steps it can make, of decisions it can make. So it's not prescripted to a linear workflow, but it can kind of make an open-ended set of choices. Um, still, you know, I like to think of agents as being a little bit distinct from the next level of agents, which I'm using the term claw here because OpenClaw obviously has, has popularized, uh, this concept. But, you know, there's a market shift that I see from

agents that were still mostly reactive, um, you know, to user input. So you would have an agent go and perform a task. It might spend twenty minutes, maybe up to an hour to perform the task, and then it completes. Claws, um, as I call them, are really, you know, about agents that go and perform work for much longer, right? They can even have a mechanism like the heartbeat mechanism within OpenClaw, um, and now in other products like HyperAgent, that enable them to wake up on their own and almost like, um, exhibit this always-on behavior. I mean, underneath the hood,

what's happening is, you know, each wake-up basically results in many different turns, so the agent loops upon itself. It's performing tool calls. Each tool call response then invokes the, the LLM again until it reaches a stopping point. But then that stopping point gets reinvoked or, or kind of restarted with a heartbeat mechanism where you can wake up the agent again and again, right? And so this idea of this always-on agent that's actually constantly moving towards its goal and kinda figure out creatively how to unblock itself, where maybe prior agents would

get stuck, I think that's a really, really important kind of next form factor, uh, leap that we're seeing. And then finally, as you think about models that are now capable of, you know, really, really intelligent orchestration of complex tasks, right? Um, so I've personally been using the, uh, workflows capability in, uh, Claude Code quite a bit, uh, where it can fan out a lot of work to a massively parallel set of, uh, of agents. Um, but this concept of, you know, this agent that can go and orchestrate with other agents, right? An agent-to-agent interaction,

um, to perform even more advanced tasks, uh, right, longer running jobs, is, I think, the next leap yet. All of this is enabled because the models are getting smarter and smarter to the point where they can achieve coherence even on longer and more open-ended scoped problems, right? Um, if you recall back to the days of, um- Uh, BabyAGI and AutoGPT, those were really kind of fun, cool science experiments, but ultimately, you know, as much as they were really interesting to watch and, you know, see them work for hours and hours on a

problem, they would all inevitably, you know, kinda end up drifting off to something not super useful, right? It would go off into a rabbit hole and kinda get stuck. And I think now the models are actually smart enough to enable not only longer running jobs, but actually enable us to start applying these kind of organizational principles to break down problems into differently scoped tasks that different agents can hand off to each other, all of which is about managing each agent's task, uh, or context window and ability

to coordinate work with each other. So all this is kind of abstract. I wanna make it very, very tangible and show you how we've applied some of that design thinking of building a horizontal product and distilling a lot of the complexity of agents like we did for building a database or a database-powered app with Airtable for agents in HyperAgent. So this is a, um, a very realistic use case based on some of our actual customer usage. One of the interesting things we found from launching HyperAgent is, you know, the, the, uh,

prototypical users for HyperAgent are not who you would think they are, right? I mean, we definitely get our fair share of, like, the super AI forward, you know, kind of pure software companies, like AI companies themselves, but we also get really interesting other companies that are, you know, sometimes like traditionally offline businesses. And in this case, we're depicting a very realistic scenario, which is, um, a customer, uh, that is in a landscaping business and actually uses agents to run virtually end to end every part of their business, which is finding

clients to propose landscaping projects to, like physically going and, like redoing their garden or their backyard and end to end managing everything from the prospecting to building a quote and a pitch to managing the work itself. And I think this is a huge, huge, um, interesting, um, insight that, you know, I've just kind of stumbled on from working on HyperAgent, which is ingenuity of agent builders is not contained to just AI companies or software companies. Um, and I was really inspired by Gary's talk just a second ago. I think the, the, uh,

the motive behind that applies to every industry, every sector. And there are really, really creative people, many of whom are in this room, who can go out and disrupt your own industry, whatever it be. You know, it doesn't have to be a pure software business. Um, but I think all it takes is the ability to imagine what work could look like when you do go and become fully, not just an AI native company, but I like to use the term, you know, a fully fleet of agents native company, where you're actually going above and beyond and not just thinking about small kind of narrow steps to automate with

AI, but actually how to employ a whole fleet of agents that have governed behavior and have the right context and the right tools to be able to do really useful and very end-to-end autonomous work. So let's walk through what this looks like. Um, this is a realistic example of, you know, in this business, this, uh, this landscaper wants to go and pitch clients, right? So, you know, they get all these inbound, uh, submissions of clients who are interested. Somebody submits this inquiry saying, "Hey, here's my current backyard. You know, it's a little

messy. Here's my info. Like, can you give me a quote and show me a proposal of what you could do?" This basically gets routed into a, um, in, into one agent that is the triager agent, right? And so this agent is basically taking in this intake and saying, "Hey, like, what is the, um, the scope of the project? You know, like, is this something we wanna take on? And is this a quality lead? Is this a legit person?" And so it does some research, but ultimately it hands off then the work to another agent, a surveyor

agent, to actually go and then look at the, um, the physical landscape of this person's backyard and come up with a proposal of what a re-landscaped version of that could look like. So here, um, the surveyor agent is going and looking at the original imagery, uh, or even video that the, um, submitter gave them. Um, HyperAgent, uh, in our case, uh, works with a fully, um, capable sandbox VM. So every agent, in fact, every thread with an agent is able to fully write

code, it can manipulate files, it can actually use things like FFmpeg to like, you know, clip out individual screenshots from a video file and then analyze those, um, and actually go through and, like find all the relevant information to create a fairly high quality, um, and high touch, uh, proposal that would've been unfathomable for a small business like this before, right? If you think about like a really high-end interior designer or maybe like a very high-end landscaper, maybe they could do a very bespoke proposal like this for like a

five million dollar plus, uh, client, you know, if they're working for a big hotel or something. But it would've been unheard of to do something this high touch and actually mock up real imagery and a real client pitch deck for, you know, just a small backyard landscaping proposal. But this person has done it because they've, they've figured out how to use agents in an extremely capable and autonomous way to do the job that otherwise would've been undoable by humans. And so what happens next is, you know, the surveyor agent, it, it went and created this proposal and

pitch, and now the intervention has to go back to the human. Like, there is a step where humans still need to unblock the agents, right? Sarah Guo put out a great post, uh, a few weeks ago about, you know, just describing like where the value is going to accrue in this AI landscape as things evolve and models get smarter, and I firmly agree with her take, which was, you know, the ultimate blocker to still be able to deploy useful agents, there's still a human element to that. You know, as much as models are getting smarter, the ability for agents to get unblocked by humans that still

own policies and decisions and really importantly, just kind of gate, you know, really critical, um, high impact actions, like in this case, actually making a pitch to a customer that has a binding quote, um, or perhaps like agreeing to do the work, that becomes something that the human then intervenes at, just like a great manager who's getting upwards, uh, reports or, you know, upwards work updates from their team to then unblock. So here we have Sage who has come back to, uh, the real human, uh, owner of this business and

saying, "Hey, here's a lead. I already surveyed the, the project. Here's a beautiful pitch deck and a, um, proposal, and have come up with a realistic quote based on all the factors I've already surveyed in this, uh, particular project to give to, uh, to you to now approve sending to the client and see if they accept it," right? And you can see like even the, um, the proposal itself is this really beautiful kind of interactive, uh, webpage that again, would've been unfathomable for a business like this to be able to send to every single client, even for

small tasks, uh, or jobs like a ten thousand dollar landscaping proposal, right? Um- You know, better yet, as you interact with these agents, you know, as, as Gary talked about with gBrain and, and just kind of like creating this very, um, sticky context layer that learns with every interaction, I completely agree. I think that is a, you know, kind of a universal principle and kind of a form factor for useful agents going forward is agents should not be static. They should not just be, you know, either static LLM calls or static even agents that

just have a prompt and, you know, some static skills. They really need to learn with every interaction, right? And you can't wait for the next, you know, kind of fine-tuning run of your own models where you have, you know, a bunch of collected data or rubrics to, to retrain the model. It should learn in real time, right? It should accumulate the memories and skill updates and just take user feedback and actually learn from real-world performance to constantly adjust and get better just like humans do. And so in this case, you can see, you know, within Slack or, you know, in our case within, uh,

either Slack or email or other, um, invocation methods, you can basically go back and forth with the agent just like a real human and say, "Hey, like, you know, remember to take this into account next time." You kind of, you know, here's something differently, uh, formatted or a different way to think about the proposal, or you missed kind of, um, this additional cost lever in this particular project. So, you know, remember that for next time. And I think it's critical that in this next generation of agents, agents need to remember these things, right? Um, they should feel fluid and, and,

um, evolving just like humans are in order for them to fully realize the potential that the latest frontier model capabilities enable them to, to actually go and do in terms of autonomy.

Um, and then finally, I think what ends up happening at the end of all of this is with all of these agents that are each individually capable and even interacting with each other, I think the job of the human becomes increasingly about unblocking agents to be able to do work effectively. And so, you know, I, I compare this to the development workflows that, uh, have evolved from, you know, originally before any, um, you know, kind of AI en-enabled coding, obviously, we just sat in front of a computer. We as humans are pretty single-threaded, right? You'd stare at one code file, you think about a problem, you'd kind

of bang out some code, maybe take a break, come back to it. To then we had, um, augmented coding with the first generation of GitHub Copilot and similar tools where, I mean, I think of that as really just a completions form factor on coding. You type some code, it can auto-complete a few extra lines. To then we had things like Cursor Composer 1.0, like the original, uh, experience, not the, the composer model, um, but the chat experience where you could talk to this agent, it would do more complex changes, and that was even more leverage, right? But we're now moving to this point where, you know,

the best frontier agentic developers I know, and many of you are sitting in this room, are, are really just kind of overseeing a fleet of agents, right? Like, in fact, going to sleep without setting off your agents to perform useful work overnight feels like you're, you're taking this huge loss, like your, your team is not working because you're not unblocking them. And so the UX of how you oversee these agents, whether it's in development workflows or general purpose work, like this landscaping example, needs to shift as well. And so the way we've thought about that is you need to see

this kind of orchestration control plane where you can see at a glance what all of your agents are working on, what they're blocked on, who's handing off work to each other. And your job becomes almost like, you know, zooming out to see the entire Sim City landscape and orchestrating across everything going on at a macro level between your different agents rather than going and having to play micro and zoom into each individual project or each individual task. Um, and so I think this is a really, really exciting moment as we have shifted

from, you know, not just completions and kind of the slight augmentation of existing human, uh, work, but really a radical now leap into humans as orchestrators of agents that also can orchestrate themselves. And finding the right way for us to interact with those agents as they become more and more capable is gonna become not just the opportunity to boil the o-ocean, but in my opinion, table stakes for survival for every company in every industry. This is gonna be the way to get ahead and thrive in your

industry. And if you don't, you know, as Jensen put it, you know, it's not gonna be AI taking your job, it's gonna be, you know, somebody using AI who takes your job or takes your business. So really, really excited both for the opportunity and for everyone in this room to be part of this, um, transformation and this really kind of bold leap into the next form factor of AI. Um, all that being said, we'd love to see you try HyperAgent, so hyperagent.com/aie and you get $1,000 of inference credit, um, you know, that can be spent on all of our different model offerings, including, uh, you know, the

latest, uh, Opus 4.8, Fable 5. You can use it on GLM 5.2, uh, on GBD 5.5, et cetera. And we'd love to see what you do with, uh, with agents on HyperAgent or elsewhere. Thank you all for being here.

We love Howie in here.

Well.

Thank you for joining us.

Awesome. Thank you.

Um, so w- for the close of our events, uh, a few w- months ago, I saw that HyperAgent was launching this startup competition. And at AIE, we've always wanted to feature some kind of battlefield for like a com-competition and a contest for people. Uh, and we didn't really have any sort of mechanism by which to judge or to fund it or anything like that. And HyperAgent came along, and they had this, uh, 500 Founders thing.

Yep.

Right? W- what was the inspiration for that?

You know, I think, um- As I think about the multiple layers of disruption stacking, you know, there-- we, we're going, as I said, from, you know, just humans doing the same work they were augmented a little bit by AI to now humans overseeing fleets of agents doing work that wasn't before possible, like this landscaper running his business in a fundamentally different way, to now this entirely new economy of agent-native companies being built. And so, you know, really the Founding Five Hundred is about betting on the most frontier part of that economy as possible, right? It's the most agentically leveraged companies that are not only out disrupting their

space, but kind of defining an entirely new playbook for how companies in this new era will be run. And I think the leverage we're already seeing from some of these, uh, these companies that are running in this way, like Gary said, like, it's profound. I mean, it's far more than even being like an early software leveraged or internet leveraged company. So really, really, you know, kind of big believer that this subset of the economy is going to have explosive growth far higher than any other sector, and it's gonna cut across not just, you know, the tech industry, software industry, but every industry.

Yeah. Uh, in my London keynote, I always-- I often talk, talk about how, uh, work, uh, agents are eating the rest of work-

Yeah

... and being accessible to non-technical teams. In fact, my team is right there back now because AIE runs on Airtable, uh, and they're so excited to meet you and so excited to-

Oh

... to work with HyperAgents. Um, so from five hundred founders, we-- you guys preselected twenty. They competed today in the Startup Battlefields. Uh, three have emerged as finalists. Let's introduce them. Um, so I don't know if I'm supposed to, to click this. Uh, but, uh, first of all, I'm just gonna go to introduce, uh, our dear judges. I think, uh, we're gonna have, uh, Howie, Theo, and Joshua. Um, thank you from Agent. Um, and, uh, we'll-- I think we'll, we'll, we'll sort of,

uh, have them, uh, judging there. Uh, but also we're gonna, uh, go into the format and the background as well. Uh, so basically what we're gonna do, uh, it, we, um, the HyperAgent team has kindly put up a hundred thousand dollars in prizes and basically what this is, is a competition for whoever is gonna be the most impressive, uh, winner. So, uh, fifty K to the winners, um, thirty K to runner-up, twenty, twenty K to second runner-up. Uh, we're gonna have a hype video that is, uh, h-thankfully,

um, created by Hyperframes and Agent. Uh, we have founder pitches and Q&A with the judges, and then we'll also get to the next, uh, s-- uh, pitch as well. So just three quick pitches back to back. Um, and basically we decided that we really wanted some kind of judging in terms of like what is impressive and usage of this technology and like why it matters. Um, it's not an investment pitch so much as like, like, you know, it's not like a VC pitch so much as like, well, why should you

care? Why is this an interesting use of technology? And, and why, um, you know, why are you like an, uh, interesting founder that people should know? Um, and so I'm very excited to announce like the winning teams. Uh, so I think first of all, we're gonna introduce, um, Commod. Spiro, come on stage. All right. Thank you.

Um, I think you guys are gonna take over and present, yeah? Good to see you.

Good to see you.

Hey. All right. Do you have your clickers or y-you got, you got your...

I don't know.

You're gonna just kick it off? Okay. All right.

Hey, everyone. I'm Spiro Nanacos, one of the founders of Commod. This is Dom. And I'm Nick Nanacos. We are the founders of Commod. Commod has agentified physical commodity trade. From the rice you eat, to the gold chain you're wearing, to the chip in your smartphone, to the gas you're putting in your car, to every single copper pipe running through our data centers, physical commodities power the world. But the global trade system is broken.

Trillions move annually, yet hundreds of billions are lost annually due to fraud, fragmentation, and the sheer complexity of coordinating global trade. Commod has taken that complexity and transformed it into an agentic execution layer. Our proprietary CFC state machine governs specialized agents that verify, execute, and settle trades entirely. And then, you know, people ask, "Why is this so complicated?" And if you think

about it, it's like, well, it's every country in the world, every resource-

We should-

... all of their different currencies and all of their different banking laws.

We should end the video.

So of course, it's inevitably going to be a disaster, right? But with AI, it's finally solvable.

Okay.

So we're very excited to have you here. We also have our, our first customer in the crowd out of the UAE, um, trading company based in Dubai, and he moves several hundreds of millions of volume in, in trade every single year and, uh, he's proved the product. So we're excited to be here today, and thank you for the opportunity.

Talk to the judges.

Seventeen trillion dollars in physical commodities move every year. The world runs on them, yet it still moves on paper, email, and a handshake.

Okay, so we'll do the next one.

No one knows who is real or if the documents are.

Got this one.

The result, billions lost.

Which max? Okay, max.

Commod plugs into the stack you already run.

Interesting. Okay, yeah, yeah.

Gmail, Slack, Stripe-

It says founder one down there

... Salesforce, WhatsApp, and more. A deal begins. Agents screen both sides. KYC, identity-

Max

... sanctions in seconds.

Max. Okay.

Forensics catch document fraud instantly. A private deal room spins up. Shipping and logistics partners link in with live pricing. Then an agent swarm structures the deal in parallel. Compliance, legal, currency, banking, shipping, settlement.

I think max is next.

Weeks of work in minutes.

It's max.

The swarm obeys one protocol, the CFC, our patent-filed state machine. Agents advance only by clearing every deny condition. It cannot move until the evidence is real. Settlement eligible. It never touches the funds. It emits the signal. Your bank moves the money. Thirty days closed in minutes.

Guys-

Commod, where commodities meet trust

... the judges can't see the video on the confidence monitor.

What happens now?

I don't know. Is it on me?

Yes. Um, I, I, I realized that we didn't get you a mic, so, uh, are you... You're still mic'd up, but the other two-

Yeah, yeah, yeah.

We're all mic'd, yeah.

Oh, you're all mic'd up. Okay, great.

Yeah.

Um, so I think we have a little bit of a Q&A session.

Yeah.

So are we supposed to ask questions or what?

Exactly.

My team agreed to this, not me.

Wait, why are you guys playing this?

Joshua, you wanna go?

Uh, sure. I, I'm very interested in the, um, the second project, Comment.io. And, you know, certainly we need a shared space for agent to collaborate together. Just curious, have you thought about having that format to be HTML versus like Markdown file? How you been thinking about it?

What is this music?

Uh, sorry.

Yeah.

All right, sure.

Yeah, don't-

Yeah. I was just saying that, you know, um, um, the second project-

It's true

... Comment.io i- is about-

Yeah.

That's the next one, yeah.

Oh, the next one.

Yeah. We're asking them about the first one. Uh, so I'll ask a question. This is one of those types of products that's really hard to build, 'cause it requires so much domain-specific knowledge, as well as, like, the building out of all of these pieces to collect the information needed for the product. A lot of the concern I have with modern startups is the thing that they're building is eventually going to be a, a feature of a sufficiently intelligent model. Do you see a future where the models get capable enough and the tools they're given by Anthropic and OpenAI could compete with you guys if they were

to get smart enough and build in the right directions?

You want me to take it?

Yeah.

So in commodities trade, it's very specific as to what data's being pulled, and what we've done with our orchestration layer is set the gates for each agent. So they're pulling from real databases and have that context themselves. Whereas you're talking about gold, oil, grain, whatever it may be, there's certain procedures that only industry experts really know and have the documents and the information to pull that in, which is why we've partnered with Connor and a few other partners to bring in all that data, and I have domain experience as well, to know what's real, what's not, and to

train the agents based upon that. So to your question, could that happen? They could learn more for sure on the open internet, but this is a private industry with a lot of sensitive information, and they have to learn upon and train upon certain data in order to progress and become experts in that domain.

So it sounds, like, relatively bespoke in the sense that, like, every company's a bit different. You need to get everything for them. How many customers would it take for this to become a billion-dollar company or to hit, like, I don't know, 100 mil ARR? Like, how many customers do you think it would take to get there?

It's about 100 com- uh, 100 customers total.

Huge.

It's also important to mention as well, like the tickets in this industry are massive-

Mm-hmm

... on, on deal sizes. It's not like a Stripe, for example, processing micro transactions. One transaction might be, you know, $200 million or north of that.

And one partner such as a refinery, many of them transact $9 to $12 billion a month. So getting in at that level, charging per deal on that changes the whole landscape of what we're able to earn and what we're able to be valued at.

Yeah, just wanna add one more piece as well. Our solution is end to end, so you know, the solution serves the entire supply chain, but the wedge ultimately into the industry is the traders and the banks.

Hmm.

Banks pulled out of physical commodities trading years ago after the fraud became so rampant. So this unlocks a tremendous amount of liquidity as well by the audit trails and the immutability that we give, uh, into the commodity and the trade itself, so.

There's a lot to love about this business. I think, like, in many ways it's, like, check, checks all the boxes on, like, the classic, like, here is the ideal vertical agent business to build. Like, if you look at the w- the YC kinda, like, RFS for vertical agents, like, this is, like, the perfect application in the right, the right kinda sector. Like, there's lots of money up for grabs, and there's a lot of, like, domain-specific stuff that you have to do that creates, I think, a good defensible moat. Um, I think my perspective is more like, you know, what do you do to keep up with the changing competitive landscape and just, you know, as

models progress, how do you need to evolve your product to stay competitive, right? So it's less about is this a relevant and great business today, and more about what changes over the next two, three years or even, like, next year that impacts you in a way that... You know, it doesn't necessarily mean, like, Anthropic and OpenAI eats your lunch, but what do you need to do to stay, you know, kind of at the frontier and be a better product?

Yeah. I think, like, the only thing that's actually defensible is being the most creative in your category, regardless of which product you're creating, right? Like, truly nothing's defensible other than being the most creative, right?

Yeah.

Anybody these days has a product.

Yeah.

But we actually do like the Anthropic eat your lunch model. Like, we like the ship extinction model where we ship and then 10 companies go extinct the next day. So, um, yeah, it's aggressive shipping and product superiority, but then ultimately we're gonna win on distribution.

What are you most excited about on your, uh, upcoming roadmap, if you can share?

Becoming the bank.

Yeah. It's exciting.

Thank you so much.

Awesome. Thank you.

Awesome.

Give it up for Commod. Uh, we're gonna invite the next team on now, uh, if you wanna come back here.

No worries.

Thanks.

Thank you.

Yeah.

Appreciate it.

Great job.

Yeah. It's awesome.

Thanks.

Yeah.

Um, and up next we have Max Winderbaum and the team from Comment.io. Come on, Max. Woo. Oh, you're gonna play the video? All right. Come on in.

All right. We'll play the video.

That was- That's the music.

Oh. That's all right though.

All right, welcome in. Come on, guys. Oh, thank you.

Um, if you could, uh, so we have a little bit of a technical snafu where the judges couldn't see the video.

Okay.

Uh, we have to describe a little bit just to, just to prompt things off.

Before.

If you sort of, if you sort of re- before you get into the whole pitch, yeah.

Oh, okay. Um-

Pretend like we're a non-multimodal model, and we just, we, we only accept text.

Just giving you, giving you all the context here.

LM5 too.

Yeah, yeah.

So, uh, yeah, Textio, um, um, uh, Comment.io is a, uh, it's um, a multiplayer markdown editor. It's, uh, focused on making the best documents, um, for the modern work, which is, uh, people and agents. But what you would have seen on the video is a familiar document editor interface, but people and agents are popping in, they're notifying each other. Everyone can call anyone in, um, and that's, that's basically the gist.

Awesome.

All right. So, um,

it's Friday afternoon, ready to go home, packing up your laptop, and all of a sudden you're hearing dings around the office. Server's down, big outage. So you put down your bag, you head into the conference room, and you're pulling out your laptop. There's one question on your mind. What is going on, Claude? And you're entering that in, you're waiting a few minutes. Give me some ideas. It's culminating or whatever. And then it gives you some ideas, and the first thing you do is you tell everyone else in the room

the ideas, so they can put it into their Claudes. And then someone else has a different idea, and then they let you know, and then you put it into your Claude. And you real-- How did we get here? How is this the job? There has to be a better way, and there is. It's agents and the pe- who are doing the work, and people who own the problem sharing context in real time. But how? Think of the productivity software that you all use, the collaboration software. Can you just, like, drop your coding agents into them and have them working with each other?

No, you can't because they were built for people, not for agents, and that's why we're making Comment. It is the multiplayer markdown editor. It is-- we're laser-focused on making it the best document platform for modern work for people and for agents. It's open, it's pluggable. That's key for enterprises. It's key for y- it's key for developers, and it's wrapped in a user experience that is super familiar. I started my career in productivity, bringing collaboration features into

Microsoft Word. Max and I worked together, um, at Textio, the first AI-native writing software six years before ChatGPT. We were deploying company-wide rollouts of AI writing software to the Fortune 500. We've been doing this a long time. The market for Comment is expanding with agents. As more and more people use agents, more people are going to need the ability for agents to share context in a product that is built for agents, but designed for people

working together, even in the highest-stakes circumstances.

Awesome. I love the shape of this business because it's like the inverse of what we just heard from the, um, the Commod guys, right? Like, they went super vertically deep. They're trying to extract a lot of value out of a very deep, um, you know, domain. And you guys are going super broad and arguably maybe a little thinner, but by intention, right? And I don't say that as a, as a detriment at all. Like, I think, you know, one way to think about the landscape here is, like, the agent economy is gonna be so large. There's just so much money flowing through agents, like literally the token spend, right? And if you can be

a product that like even a one to, you know, two percent, uh, amount of those tokens touches-

Mm-hmm

... like you could be a very valuable-- even if like the per, uh, per call or per usage, um, you know, kind of, uh, metric, uh, you know, kind of margin extraction or revenue is, is relatively low, you can be a very cheap, very fast, very, um, broad product and still gain a lot of scale, right? So I think it's a really, really interesting shape of product. Um, what have you found is like the most surprising learnings from like a product design or product affordances standpoint to make this product better for agents than like the human-designed equivalents, right? Like, why not just have agents collaborate on a Google

Doc, right? Or even try to use like GitHub as the, uh, you know, kind of repo, not just for code, but for markdown.

All writing software today is horrible for agents to use. Software like Google Docs has a really terrible format. It has an MCP. You can send a document there. You can have your agent read a document from there, but you can't have them collaborate in real time, and that's critical for all kinds of real work that needs to happen. And as more agents join the wor-workforce, they need to be in there with your documents. And as we've talked to

our users, people who love working in documents, who rely on documents for their day-to-day, need to bring their agents because they're starting to use agents too.

Yeah.

It's not just developers anymore.

Okay. I think-

I'll mention it-

Oh, final closing words.

I was just gonna say, the other thing I'll mention is that I've been super surprised about, um, talking to people at this conference when we've talked to them, is how many people have said, "I've rolled my own like visual markdown editor that other people can look at," so people can look and then you're like, "Oh, but I gave up. I just ran an HTML document instead." Like, it's a hard problem, and you need dedicated, like, focus on making it great, and that's kind of what we're doing.

Awesome.

All right. Nice job, guys. Thank you so much.

Uh, next we're, and last team to compete, we're gonna play the video by Foundry.

Content creators don't have time to build a business, so they have to hire an agency. It's expensive, slow, and risky. Built By Foundry replaces the agency and lets creators launch their own products faster and cheaper. Our AI agents build the business end to end. The creator launches it. We keep growing the revenue. Type in any creator's handle, and our agents become their biggest fans, watching their videos, reading their comments, and

finding a painkiller their audience would actually pay for. We've done this for creators of every size, and over half a million people use products our agents built. The cost of building software is approaching zero. What's left is trust and taste, and no one has more of it than creators. We're building the agentic product team that is going to turn every creator into a founder.

Woo! Foundry team, come on up.

Thank you. Thanks, Max.

Hello, AIE. I am Weston Bell-Geddes. I'm the founder and CEO of Built By Foundry. We're gonna turn every content creator into a founder of their very own company. And I have great news today. I think we found the formula for a winning business. You take your unique expertise and turn it into a product, and you solve cold start with your built-in distribution. Theo, you, you did this with T3 Chat.

Yeah. It sucked. It doesn't work.

It works great for our, our, our, our, uh, users, but, uh, Gary did it with, with G s- uh, G Stack and G Brain. And, uh, what, what we, uh, find is that if we help, uh, creators to take a step back and solve a real pain point that their users have, so we're not just making another Patreon for that, paywalling this content. We're helping them solve a real pain point behind that, and that's what our agents are experts in, is finding the pain point within the user's problems.

And our companies have, uh, b- built over, uh, many different niches and, uh, different sizes of sha- and shapes of creators, and really what we're building is we want to give the power back to the content creator. They're stuck in these... I worked at TikTok for two years and, and I met with hundreds of creators who are stuck just doing brand deals or affiliate links or selling $20 PDFs in their link in bio, and they are one step away from building their own

business. They just need someone to help them, and that's what our agents do. At Built By Foundry, we deploy teams of agents to build recurring revenue businesses for content creators so that they can focus on posting and posting and posting, and we can keep growing their revenue.

I have a lot of thoughts on this one.

Okay.

So the problem I find as a creator, even with the brands I work with-

Mm-hmm

... is that there is a saturation point where it stops being useful to talk about a given product with my audience.

Mm-hmm.

Like with me for Vercel, for example, there was quickly a point where everybody in my audience either was a Vercel user and doesn't want me to shill it to them anymore, or they won't be a Vercel user and they don't want me to talk about it anymore.

Mm-hmm.

It is very easy to hit the saturation point, and the best solution is to have a much broader set of things to show your audience.

Mm-hmm.

One of the best things I ever did for myself, the single moment that has had the most positive impact on myself, my career, my businesses, and my income, has been going from having four sponsors to 80. Because now there's a wider variety of things to bring my audience, and even if 80% of them aren't relevant to them, the 20% that is is now conversion I'm getting that I wouldn't have otherwise. The variety of what I bring to my audience is where I find the most value by far, and every creator I've worked with ultimately has come to a similar conclusion. That said, you're touching on really important

pieces here. Creators do have more taste than most founders. Creators do have better distribution than most companies. Creators know their audience better than most businesses. And you've built a machine that identifies that, but it doesn't provide the solution to that other problem, which is you have to bring variety to your audience. Where I could see this working really well is as a tool that businesses use when they identify a creator as a- overlapping with the archetype of customer they're looking for to help that business build better things for Theo. And I've told

this to so many businesses that want to work with me, "Watch my streams, watch my videos, find the problems I have, and then show me how your product solves them." That's a way better pitch than going to try and start your own company as a YouTuber. As the YouTuber who has started the second most companies, as far as I'm aware, it doesn't work. It just burns you out, and it burns your audience out too. The best thing you can do is work with companies to teach them what your audience wants and for them to come to you with money to show it to those audience, or those audience members.

I understand. And I will say your niche as a tech creator is one that is very challenging because there's a lot of options. Um, and what we have found is that for all of our creators, I'll just f- a few examples. Um, one, one creator's name is Brandon, and he does homesteading content, and he always loses track of, uh, the best settings on his freeze dryer, which sounds ridiculous to us, but to his two million followers, that's a real pain point. And now they pay him a recurring revenue monthly, or monthly or yearly

subscription to solve that specific pain point. So I think unfortunately, Theo, you've picked a bad niche. I'm sorry. But, um, if you want to pivot to homesteading or gardening- ... may- maybe, uh, we could, we could work together. But I, I'd say primarily whatever the user's niche is, apart from tech and maybe com- comedians, we've been able to take a step back. Our agents find a pain point that s- they're willing to pay for, and zero creators across our entire company history have ever left our, our company because

they always are profitable.

You're very convincing.

Okay, excellent. Thank... Give it up for Foundry.

Thank you, man.

Yeah. Thank you. Um, so we're gonna have a little bit of a de- deliberation moment. Uh, the judges are gonna go backstage for a couple minutes. In the meantime- Sorry? Okay. In the meantime, we've, uh, prepared this little video recapping the World's Fair. So take your time, judges.

Now, this is our biggest AI engineers event ever.

There are constantly AI new techniques for you to

try.

Thank you so much. All right. Um, I think we're inviting everyone back on stage. Give it up for the Startup Battlefield.

Um, congrats, guys. Yes. Okay. So, uh, we have some prizes that we're gonna hand out. I think, um, the folks back there are doing some checks. Um, in second runner-up place we have Commod.

Uh, Theo, I think you're presenting. Uh, Theo, yes. We can, we can- Okay. We'll take some- Okay ... pictures as well for the team. Awesome.

All right. Spoilers. Uh, next- Oh, no ... and, uh, I guess this also reveals the order. Uh, we have Foundry as runner-ups. Great pitch. All right. Max, great pitch.

All right, congrats. Awesome. All right. No secrets now for the grand prize, Comment IO. Welcome. Great video as well.

Congratulations. Well deserved.

Do you wanna also do the group photo? I think we're doing-

With all of us?

Um,

yes. Okay.

All right.

Let it be just them. They're the founders.

Winners.

We're just here to comment.

Just the winners.

Thank you, everyone.

Oh, fine. We'll sneak in our faces in. Get them.

Joshua.

Come on in.

Yes.

All right.

Thank you so much, guys.

Awesome.

You're welcome.

You had some building with HyperAgent and-

Yeah, of course.

Yeah.

Yeah.

And patient. Um, any last words that you wanted to-

We should talk about it later

... comment on? I think you guys, uh-

Yeah

... briefed a little bit. Yeah.

Yeah. I just wanna thank, uh-

Sorry about the-

Thank everyone here at AIE. We got amazing feedback-

Yeah

... um, through the whole, um, time. And, uh, HyperAgent, um, for inviting us to this. And, uh, HyperFrance-

Of course

... for helping us with the video.

Um, great standing by you

And, uh, overall, like-

Okay. So if you want-

Thank you so much. Check us out. It's free. Comment.io.