Why Domain Experts Are Winning In The Age Of AI

12:18Bryant Chou:And so it's operating in real time. I mean, it's probably doing the equivalent work of a team of three to five engineers and front-end people and probably take a week at least.

12:29Y Combinator:They're probably going to open up Cloud Code or Cursor to try to do this.

12:33Bryant Chou:But they'll do a much worse version of it.

12:35Y Combinator:Here's the moment of truth. Let's see if this is responsive. Boom. Boom. Boom. Boom. Boom. Cool. So it's responsive. You know, the fonts are showing up. Let's see if there's like some hover effects that we captured down here. Okay, cool. Some CSS hover effects. So these are all things, all these details. Last mile. You need to have an opinion about how websites should be made in order to do this. But this isn't the point, right? The point is not to just recreate something. The point is, after your website's done, what can it do to work for you? And that's where Play really shines. There's a bit of a nuance here that I don't know if everyone in the audience would know, but when you try to create something like this with any other vibe coding tool, it's not consistent.

13:25Y Combinator:And it kind of remixes and forgets all the design consistency. And that's pretty impressive to get it to do that. It's sort of the following the... Yeah, this is where we're at, I think, in the age of AI, which is, I think, you need to have a certain amount of expertise to know what to do with this boundless intelligence that's imbued in the model. And I think this is where folks with experience, folks that have spent a decade plus in this industry, they know how to create something like this because they can leverage the model's underlying capability to create something that's just world-class.

14:04Bryant Chou:Let's see. So say I'm Cursor, I released Composer. It's a new type of model. Can I just come and deploy and paste in screenshots and even my PM spec and you could make the product page for me?

14:17Y Combinator:Even better. So we have integrations with like 50 different tools. It can not only just connect to your code base, but it connects to Figma, connects to all of your analytics tools, CRM, spreadsheets. It can even draft emails on your behalf based off of who's coming to your website.

14:33Bryant Chou:Okay. This looks like a company brain for your marketing then. Exactly. So you start with the website, but that rapidly becomes sort of your company brain for how you describe your product and show it off. And it makes sense that you would start off with your homepage. I mean, I would probably do that the same thing too. It's like your homepage is sort of your face. And if the homepage doesn't have it, that's almost like the source of truth for how you discuss what you're building.

14:57Y Combinator:It really reminds me of Rippling in the early days would sound odd, but I remember like the very first thing Parker built was an offer letter generator. And I remember seeing it. You gotta start someplace. Yeah, but like it's the only insight you get is like a second time founder. So that time I was like, you're pitching this really big vision, but like you're starting with like an offer letter generator. Yeah,

15:14Bryant Chou:but everyone has offer letters, everyone's hiring people. So, but then that becomes an HR system. And then once you have HR, you have auth, and then boom, you're like literally your company app store OS. And if you think about the journey of a new employee, the very first step with a new employee is they get an offer letter. That's where everything else starts.

15:33Y Combinator:Well, the very first thing with the website is the homepage.

15:36Bryant Chou:Yeah. And then everything else on. I mean, even for a business, right? Like you might start with Stripe Atlas, but right after that, like I better make a homepage. And then it might be very cryptic, but when you actually launch it, it better say what it is.

15:49Y Combinator:Yeah. So it can not only just integrate to your GitHub, but your entire systems of record. But the really cool thing is that it's actually thinking about what to do while you're sleeping. So every single night, we look at all the traffic. We check your Google search console. We see what your pipeline looks like. And it's able to offer suggestions, right? Like, oh, you got an active target account. Oh, you've got someone engaging with your campaign. These are all sort of things that it's telling you like, and it's able to glean because instead of like checking your analytics every day, no ploy can literally just tell you everything out of the box.

16:32Y Combinator:Right. So this is just really basic analytics that we offer, but it can also tell you who's coming to the site. And I think that's really cool. Right. It's like, Oh, wow. well, someone from this company clicked on this call to action button, and now I can do something with it. I think what's pretty cool is basically you're taking this very esoteric niche thing that OpenClaw users know about with a dream cycle when it iterates and improves the skills, which I think only people that are in super long tail and in OpenClaw, which I don't think the rest of the world knows, you're doing that for everyone else.

17:05Y Combinator:Doing it for businesses, because I just think that so many businesses, they have a great product, they have a great service, but there's just so much sort of a met opportunity for these business owners, for these founders. And I'm really just here to make it easier for them to tap into it. So I wired this up to Y Combinator's Google Analytics and Google Search Console and pulled in all the data. And what was cool is that like right out of the box, without me having to do anything other than like click through a couple of OAuth flows, it was able to give me like a full SEO report with a bunch of like suggestions for the site.

17:39Y Combinator:And I was thinking about like how I would do it without ploy. And with enough effort, I could have maybe gotten Claude Code to do it, but Claude Code doesn't know how to connect to any of those things out of the box. So it'd have to figure out the APIs and have to suck in all the data and then doesn't really know how to do SEO optimization. So it would have been so much prompting and work to get that same result. Yeah. I think that's stuff that I'm surprised by because I think the best companies that I've met in YC, they are on the right side of model development. And what these models need is with a little bit of steering, with a lot of data, there's just so much alpha that you can derive.

18:15Y Combinator:So I think that is one of those examples where let's just feed the model data, structured, unstructured data, and just let it cook. How did you imbue it with an understanding of how to market websites, how to design websites? Are there some ways that you distilled lessons from Webflow? This is stuff we geek out about at Ploy. So we're in admin now. This is the behind-the-scenes code. Yeah, this is behind the scenes. The lookbook. So what you're looking at is our curation of what we believe is the frontier of web design. And this is stuff that you can't really get anywhere else. And we've used a collection of models, open AIs, ChatGPT images.

18:56Y Combinator:But then we've created 3 ,500 prompts for web designs that then Ploy takes inspiration from. So you're not going to get a website that looks exactly like this when you're using Ploy. What you're going to get is you're going to get some of the vibes of these sites. And I think that's really actually how human designers work, right? So if you go and work with the agency of freelance web designer, some of the best might come up with something incredibly bespoke and unique that no one's ever seen before. But a lot of them, you know, they get inspiration. And that's essentially what we're trying to do here with our product, which is like, hey, let's emulate how real humans work, and let's think about how we can create some really unique layouts, really unique designs that can really stand out.

19:43Bryant Chou:Ploy is basically like the anti-slop. It looks like given the corpus of data that you have, the context you give the agents, it actually sort of overwhelms their inbuilt predilections. One of the examples is like, for some reason, And the models really love that left-hand rule with the rounded corners.

20:06Y Combinator:I mean, in web design, there's so many AI tells. I would like to say that Floyd eliminates all of them, but you can't necessarily eliminate all of it with all the prompts and guardrails and steering. However, we have spent a lot of time to make sure that the sort of essence of the human individuals and like the businesses bespoke sort of representation, their brand can be reflected, employed to the best of our ability. And I think like the best analogy that I have for where we're at in the AI cycle is like Andy Warhol, you know, created paintings, but, you know, the stuff eventually ended up at a factory and the factory would use machines to recreate these prints.

20:48Y Combinator:But it's still Warhol. And I think that's where we're at, which is like these models, they are essentially the factories for human creativity. And that's essentially what I want to be able to deliver for digital marketing. Why digital marketing? Like you could have done like factories for anything. Why is this the thing you decided to build the factory for? I think that the web is still one of the most transformational technologies. Sure, it's done some funny things with how we consume media, how we respect sort of media organizations. However, ultimately, the sort of democratization, the dissemination of information for the Internet, to me, is a really exciting tenant.

21:31Y Combinator:It's a really exciting space to be, even after working in this space for dozens of years. My approach is that where we're at in AI, these businesses, they still need to be found. And I actually think there's going to be way more small business in the future. You're not going to have massive companies anymore that are dominating. I think where society is moving is I actually think entrepreneurship might be become way more important than it has been. Entrepreneurship, especially if you're a small business, you need to be found. You need to be able to tell your story. You need to be able to represent your brand well.

22:10Y Combinator:And I think that's just like a really exciting space for me.

22:13Bryant Chou:This is like 13 years after your prior company, Webflow.

22:17Y Combinator:You were my group partner, remember?

22:19Bryant Chou:Yeah. I remember you coming in with basically the best visual, like graphical user interface for building a website because it was built directly off of CSS.

22:31Y Combinator:Yeah. So, I mean, it was incredibly novel at the time, but, you know, we created this visual interface over HTML, CSS, and JavaScript. And it was sort of, you know, the first no-code application that was out there. Power's, you know, big portion of the internet today, something I'm really, really proud of. It's just one of those products and companies that I think it's just, I never knew it would become that big. You know, I remember doing office hours with PG and PG couldn't figure it out. And it was like, started to stress me out. We're all sweating.

23:04Bryant Chou:He doesn't know CSS. Not at all. He knows HTML tags. And you could like put like font properties on it, you know. But that's about, you know, Hacker News was more or less like about what he understood.

23:16Y Combinator:Yeah. And like for someone that's like as technical as PG, we thought that. He was like, oh, he would be able to get it. I mean, he built Hacker News. So this was 2013. What was the market for website builders like then? I think it was pretty competitive by that point. It was extremely competitive. I could probably, I remember there's probably eight of them, four of them I still remember. And then even in our history, there are so many different ones that popped up, right? So I think like the things that we did really well was, and it really bothered me at the time, but my co-founders, Sergi and Vlad, were like perfectionists.

23:51Y Combinator:and they just like stressed over every minute thing and when you're building something that's supposed to scream pro it's supposed to scream craft that really helped maybe go into more detail like how did you come up with the original idea and i think the reason it's interesting is it feels like from our perspective in the batches we chat gpt launched and there was just like green field for like all these ideas to build now it feels like we're entering a bit more of an age of where it's competitive and it's harder and often you're entering in a new idea, but you've got several competitors. So like, how do you build the confidence to enter a space that's competitive?

24:26Y Combinator:And how do you know that you have like an edge that's worth digging into? I think for me this time around, it's almost opposite of Webflow. At Webflow, we focused on one persona. That persona was this like freelance web designer, you know, just like, there's probably only 50 ,000 of them, honestly. With Ploy, you know, we're essentially solving for tens of millions of people. And I think that's just like a really interesting thing now that you can do now with AI. I mean, we've talked about like boiling the lakes, the oceans, like this is a very much, ploy is very much a boil the ocean.

25:01Bryant Chou:You couldn't do this before, but now you can do it. And you can do it to just sort of an award-winning degree.

25:07Y Combinator:You know, I've been in tech since 2006, like 20 years. And, you know, when cloud computing came out, it was like, oh, this is revolutionary. I've got like untapped sort of compute, networking, storage. And now, you know, as of a couple of years ago, there's this new primitive of intelligence. It was just like so irresistible to not build something in this space. How has that changed the way that you've built Ploy? How was it like the first three months or so of Ploy being compared to the first three months of Webflow? To be honest, when we were in YC 2013, we worked a lot. And we also covered so much product in those three months, manually coding.

25:48Y Combinator:Like, what is this? I have to type stuff? So I would say that we covered a lot of functionality, just coding. But now it's obviously a different league. Everyone can code. You don't have to have all this background and infrastructure and systems design. These models are really, really good at it. So I would say the biggest difference is the output. You're probably going to get more tests than you ever wanted. You're probably going to get way more code coverage, way more sort of functionality. But the thing that hasn't changed is what to focus on and how to actually mold it. I think that's something that still benefits experienced builders.

26:30Y Combinator:That's something that if you have a very strong vision, if you have all that track record of building products, that's where AI can really help you. I'm curious how that factored into building ploy. Like you've been doing website design the old fashioned way since 2012. You probably watched a million people use Webflow to design a website. Can you like remember decisions you made and how you design ploy that maybe somebody who didn't have that experience would have like not done it that way? I have such a great example of this. For example, I just like have spent so much time in a visual builder, right?

27:02Y Combinator:I was like, You need this panel to help you like drag and drop and to help you resize elements, to help you control the flow. And we stressed a lot about how to like bring some of that visual tooling into Ploy. And then we just kept deferring it and kept deferring it. And we're essentially at a position now where if you essentially just give the models enough context, screenshots, images, and just essentially here's something that you can do right here is you can essentially just use our annotation feature and say, click on this, and it's like rewrite this copy, make it super bold, and you just send it off.

27:42Y Combinator:And now you can just have this web designer that's imbued in the product that just sort of absorbs your intention and translates it into incredible outputs.

27:54Bryant Chou:Let's play it out a little bit. The models are going to get, I don't know, unbelievably better from here. and then where does that leave ploy? Like, you know, how good is it going to get from here? What's the future of the web?

28:07Y Combinator:I can't predict the future. However, I can predict that there will be tons of businesses out there that are going to want an opinionated solution to help them solve real problems for the things that really matter to them. The way I think about it is that these underlying models, they're really good at a lot of different things.

28:29Bryant Chou:Their general purpose.

28:30Y Combinator:Their general purpose. And I think there's just going to be a big need for something that is purpose-built to help a customer achieve an outcome. And that's where products, even pure SaaS products, still have a right to really kind of explore that and to really leverage the model capabilities to benefit the end customers.

28:49Bryant Chou:I guess there's an interesting competitive play here. Like, basically, if ploy becomes really, really good, you know, one of the things you do, for instance, is everyone's on the Internet is talking about loops. But, you know, you have basically a marketing and SEO slash GEO marketing loop. You build content. You figure out how to say what you need to say to get someone to understand what it is to want it. And then you improve it. And agents see it. Agents want to use it. People see it. you know they click on it they want to use it in the future like if you're not using things like ploy then your competitors will win so there is like a competitive dynamic to it which has already happened for coding tools for instance like sort of unconscionable in uh 2026 to not be using cloud code or codex or cursor you know just you wouldn't you wouldn't be able to stay on top of what's going on.

29:47Y Combinator:I happen to think that software engineers are one of like the worst customers to sell to. And they can change tools on a whim. Something new comes out over here, they'll adopt it. So I see it as like almost this market that's always incredibly competitive, but also like lowest common denominator. It's like, who's going to provide an engineer the most tokens? and that can always shift. And that battle is just really, really difficult. For me, I'm always someone, and maybe this is where my background working at Intuit sort of comes into play. It's just like, hey, just go and pick a customer that has like a true, true, true pain point and just really, really focus on that.

30:32Y Combinator:To me, it's businesses, it's small businesses, it's startups. And I think like by building this product to solve for their most important pain points, I think that's going to be one of those things that I'm always going to try to focus on. I think there's a world, which I think it is true, where there's still a lot that you can value, that you can build on top of the models because these are so good. And I think the limiting factor is knowing how to prompt them and sort of imbuing all these 20 years plus experience that you have to really get the models to focus, it is almost like you're building a very special harness.

31:13Y Combinator:We talk about a hardness as this being this thin layer to getting the models to do the right outcome. And Anthropic did it wonderfully for clock code. It took the world by storm. And within just a year, there's a lot more that can be done for lots of domains. You're kind of doing that for website creation.

31:32Bryant Chou:Yeah, I mean, you're selling skills, but you're also selling code. And it's fat skills, fat code.

31:37Y Combinator:Some of these foundational primitives. So for example, you're getting a database out of a box that is very opinionated around website use cases, CRM use cases. And sure, you can go and, you know, go use a Postgres hosted Postgres server. But if you're a small business owner, if you're a CMO, you don't want to worry about that. You're not going to go and stitch together this MCP and have your Cloud Code instance and then make sure it's running all the time. These are just, to your point, Diana, the harnesses that we're going to make sure is always on, always working for you. And ultimately, it's going to drive you outcomes without you losing sleep over it.

32:17Bryant Chou:Has anyone asked you yet for ploy, but make it really, really easy for agents to use a given product? Like, do you have a bunch of lookbook stuff around proper LLMs.txt and things like that? I mean, in the YC batch, we're seeing this wave of, For Git, it's like code.storage, for instance, it's doing super well, right? And then intsforge in the current YC batch is like sort of AWS, agent phone, agent mail, like resend. If the agents choose you, that's actually big and you're going to win.

32:48Y Combinator:Yeah, that is huge. And I think that's one of the biggest benefits of doing YC again is that I'm living two years into the future, like right now here at YC. and just being able to see these founders think in this way, I was just like, oh, I never even thought of that. And to that point, you know, you're getting a lot of stuff out of the box employee to make sure that it's AEO, you get FAQ sections, you get structured sort of schema markups, bots are crawling it, connections to all the things. But I think one of the most exciting things is being able to let an agent sign up for play. Oh, cool.

33:25Y Combinator:So that works. working on it, working on it. So then, you know, just wire it up to your CLAW. And like if CLAW needs to go and build a really awesome site, employee is one of the places where hopefully it can go.

33:37Bryant Chou:Yeah. Are you going to do MCP or how are you going to implement it?

33:40Y Combinator:I think we're going to do a CLI with skills. So an MCP would be really good if we had more constrained sort of things. However, I think with the number of things that you can do employee, I think the CLI is going to be the way we do it. CLI seems to be becoming the right UX for agents. This is why I think going back to Clockwork versus Cursor ended up doing so well. It's just so much more freedom being fully on the command line. Same thing with OpenClock. I mean, ASCII characters, they're really good. I'm curious to get your perspective as one of the more experienced founders in the batch. Again, we certainly saw Chatshubut launch.

34:25Y Combinator:There was a heavy snap towards young founders, sort of like intense rise of the young founders because it was only young founders that actually were using the technology and building with the tools. It seems like it's either coming back towards the middle, extending further in the other direction. I'm not sure. What's your take on it from being in the batch? And then to the extent you can share, when you look around and you see young founders, where do you see them having an edge? And where do you feel like, I don't know, I've got you. I do think as an experienced founder, there's just so many lived experiences you've had.

34:57Y Combinator:There's so much appreciation for how something gets done. And sometimes it actually can hold a founder back. It was like, oh, don't do that. It's really, really hard. I was burned by that many years ago. So I think like you have to kind of, for experienced founders, adopt a little bit more of that bravado and that risk appetite. but then also for the earlier founders should have maybe an appreciation for how some things you have to get right and not to bring it back to play again but you can't create a hundred websites and then expect google to think that hey you're an authoritative source and just creating content for content's sake is not the goal so i always bring it back to like the first principles of it right just like make sure you can tell a coherent story make sure you're providing value in the worlds and then hopefully people will notice and you'll succeed.

35:51Bryant Chou:What you get to do today is you get to draw on all these years of like all the things that went wrong and all the things that went right. And, um, I like the idea of you just go directly to the part of the idea maze that you were before. And then when you have a company that's multi-billion dollar company, like Webflow, like many, you know, how many people were in your org?

36:15Y Combinator:Um, hundreds.

36:16Bryant Chou:Yeah. I mean, basically when you have that momentum in like what the org is and what you're trying to do, like how often is it that you're in that little idea maze and you're just like trying to see around corners, like avoid monsters and like find the gold, right? And once in a while there's like this offshoot that comes off and you're like, well, like the gold's over there for sure. But man, if I could clone myself, I would like go and check out this other place and like there might be even more gold. I don't know. It's like one of those things that now with a lot of experience, you can actually do both of these things.

36:52Y Combinator:I think the key word there was clone yourself. Yeah. I have always lived in scarcity, scarcity of time, scarcity of my own capacity, mental and physical. But I mean, AI is here and I'm really replicating myself, not just in the products, not just in the technology that we're building, but also in the company. than also in some of the sort of really AI native ways that we're trying to build the company as well. And that's just like a completely different world. Can you talk about any of them? I mean, I think we're doing the stuff that everyone else in YC is doing, right? Making sure everything is recorded, making sure that your cloud code, you know, can access all of this, making sure that, you know, our systems and operations on the go-to-market side are as automated as possible.

37:39Y Combinator:So every single call gets transcribed, gets put in the CRM, proposals get automatically drafted, email follow-ups are automatically scheduled. And now we're just able to just take on way more, way faster, and still feel like we have room to think. And I think that's something that's just like a level of abundance that people don't really talk about when they're talking about AI.

38:03Bryant Chou:I mean, I just love the, you know, not only do you get to think, but you also have the agent Renacron that will surface interesting patterns and it'll think with you. My favorite thing to do is go into OpenClaw and just say like, what did you learn about me or YC or my companies in the last week? Like, what's the most surprising thing? And then when you have Fable 5, it's like actually really insightful. It's like, oh, I didn't notice that. That's really interesting. So one of the cool things, to go back to the idea maze analogy, is not only do you in your brain have it all mapped out. I mean, to talk about Parker Conrad and Prasanna again briefly about rippling.

38:44Bryant Chou:I mean, obviously, Zenefits was this huge, all the way up and then all the way down, sort of rise and fall. And it was such a no-brainer for me to fund him again because I knew that he had loaded all of that stuff into his brain. Like he knew who to sell to. You know, he sure did know the regulatory by then, you know, that's like seared into your brain at that point. And you could just go directly to that point again. And, you know, the difference between him and you, Brian, is that he still took two years and he went into a cave. And he had to write code with a team of five or ten people, literally in a basement in the mission.

39:28Bryant Chou:I'd go and visit him after the first six months was like, okay, cool. You haven't launched anything yet. No worries. And then like after a year, it was like, oh yeah, it's been a year. Like, but the demos are really impressive. And then after two years is like, oh man, like, are we going to launch? Like, you know, what's going to happen? And then of course he did launch, but he didn't launch like just this one little thing. It wasn't like a wedge. He actually launched the whole package of HR and onboarding in one go with insurance, like done right, you know, without having to scale with people and things like that.

39:59Bryant Chou:And so that was sort of the defining startup that I got to see, you know, sort of before AI really came to the fore in a way. And what is happening now is that instead, imagine, you know, Bryant, you're in this idea maze, but you don't clone yourself like once. You don't clone yourself twice. I mean, by my measure with Gbrain and Gstack and just like the code I'm releasing open source, I'm actually on track to do like four or five million lines of logical lines of code. This is like adjusted for like bloat and things like that. And so it's basically like 400 to 1 ,000 clones of myself from, you know, from 2026 right now.

40:46Bryant Chou:Right. So not only do you are you able to sort of like flash forward into exactly the right place inside the idea maze. But, you know, you have like 500 to 1000 versions of yourself with like your skills and your taste and all the imagination and the things that came with you. You know, like Parker still needed to hire five and 10 more people. And that was a process. It's like, how do I train people? How do I give them the right things? How do I give them like a few weeks and then let's do a checkpoint? Like, you know, did you do what I need or not? And then now it's like, actually, what would take a week or a month or a year?

41:25Bryant Chou:Like you literally do in, you know, minutes, hours or, you know, at the worst, like you could build a whole cathedral of like 20 to 40 ,000 lines of code. Like what would take a typical engineer an entire year? You do it in a few days, right? This is like a really, it's not like a little thing. It's actually a really big thing that's happening. So this is why this is the age of the 40-year-old solo founder. I mean, you don't have to be 40. You just have to have taste, you know? That's right.

41:53Y Combinator:It takes a while for a startup to catch fire, but I feel like I'm standing outside with the magnifying glass under the blazing sun and I'm able to focus it. And I'm able to focus all my experience, background, technical, and knowledge of the customer base, knowledge of their buying patterns, knowledge of these cycles, and just catch something with fire.

42:12Bryant Chou:Well, let's fucking go. Love it. Brian, thanks a lot for joining us. For those of you watching, if you actually want more customers, if you actually want to win, if you want to beat your competitors, I don't know why you wouldn't use Ploy.ai.

42:28Y Combinator:Thanks, guys, for having me.