Stanford CS153 Frontier Systems | The AI Native Company: How One Founder Becomes a 1000x Engineer

Yeah. Thanks for having us.

Actually, hey everyone. I'm Garry Tan. I was a Stanford class of '03. I took a lot of classes in here. I fell asleep in this lecture hall a great many times. Thank you so much for bringing me back. It's great to be back to the farm. And every time I come back to the farm, I'm sort of shocked that I get to be up here because I feel like I just blinked and I was in your seat.

And zooming out, that's actually desperately what I want for every single one of you. What we're talking about here is there's a grand shift. Like all those historical things, literally the new standards are being established right now. And there are people in this room who are actually going to be the people who establish those things. And Diana and I and the team at YC, we're hoping that we're—the SAFE was a legal instrument. What we're going to talk about today is actually code. And not just code. Markdown is code. Like literally the new—in, you know, we're going to link it all the way over to what a startup is, what people in this room are going to be spending your entire lives building the railroad for the rest of society over.

For our generation, we were building the internet and we were building mobile phones and we were building social networks and your generation is going to create the cognitive layer for all of society. And this—I mean what we're talking about is just like stuff that we're—these are our hunches even. Like you guys are going to go and actually build it. And so thank you for bringing us back. I mean Diana, do you want to introduce yourself?

And like hundreds of millions of dollars in capital. I mean it's just a different moment right now.

I mean AI is going to change the unit of production. When I was sitting in your seat I knew that I needed to raise money, I needed to hire a lot of people. This is about me learning how to create a new cult. Palantir was like that. YC, ultimately it's a religion, right? Like this is something that we believe that nobody else believes yet, right? That is still true.

All the things we're going to talk about—a team is still valuable, human beings are still valuable, but it's not going to be just humans, it's going to be humans in concert with agents, with memory and eval and a customer loop. So by the end of this talk, you're going to understand what we're talking about. Right now, it sounds like a bunch of buzzwords. We don't want this to be a bunch of buzzwords. We want you to take these ideas and actually implement them and remake society and we think you will do that.

Let's see. Yeah, in 2010—I mean, I'll tell you my personal story. In 2008, I got into YC, we raised about $4 million, I hired 10 people, we created Posterous, which is a dead simple blog platform and we sold that to Twitter 3 years later for $20 million. And honestly, I was able to create everything—all the software we made over 2 years with 10 people and all that capital—but me with a $200 a month Cloud Code Max plan. And anyone in this room could do that and it didn't take 2 years, it took about 5 days, right?

So I experienced that speed up recently. I created Gary's List and then that caused me to create Gstack. We're going to talk about what those things are, but as Diana said, we're in 2026 now and so a six-person team can hit 10 million in revenue with just the things that we're talking about today. And a lot of you already know this, so it might be a review, but for some of you, this is some astonishing good news.

So let's talk about Gstack. This is something that I discovered late last year. I saw Steve Yegge, a famous blogger and engineer—I believe he was an early Googler—he wrote that people using AI coding agents are 10x to 100x more productive as engineers using Cursor and Chat today. And then at Anthropic, they're about a thousand X as productive as Googlers were in 2005.

And I was like, what is going on? And so I had to try it. I opened Claude Code and of course I ended up writing—I'm around like a million lines of code, which is really really crazy.

Everything—let's see. You know, let's just talk about the things that you might read on the internet. These are all wrong. It's not just AI slop. Actually, yes, LLMs are very verbose and some of it is boilerplate, but when you create your own software factory, this is actually what you're fighting. This is actually what you're preventing from happening by default.

Yes, there are hallucinations. Yes, those are actually the things that we're trying to control. Can you make demo code very quickly? Yes. But how do you get it to production? Well, you actually have to get to 100% or 80 to 90% test coverage. That's actually one of the main reasons why plan-eng-review as a skill exists. That's the number one with a bullet skill that I use about 20 times a day to get to 80-90% test coverage so that I am not shipping slop. I'm shipping something that is actually literally usable and that I rely on every day in production.

This is very controversial. I've gotten in trouble over this. I apologize to people who took my trolling as serious. Is LOC gameable and something that might be not usable? Actually, yes. LOC on its own can be wrong, but on the other hand, if you have tests, the real measure of whether or not these things work is actually look down and does it work for you? Does it work for your customers? Are people actually paying? That's actually the true metric. LOC might be a garbage metric, but I might argue that in the age of—there's nothing in Claude Code or the model or the harness or Gstack or any of these things that tell the model to write as many lines of code as possible. If anything, the reverse is probably true. We're trying to write as dense and concise code as possible to serve the purpose and I think that's something that's quite important to talk about.

This is my experience. I got to 87,000 stars. My other project Gbrain is 13,000 stars. So basically for someone who was not coding at all in December of last year, I have more than 100,000 GitHub stars and about 15,000 people use it every single day. It's hundreds of thousands of skill invocations. And so I don't know, this is sort of what I'm learning.

Last year, probably before Claude 4.5 Opus 4.5 came out, about copilots. Today I think we're really talking about a software factory. And so if you use Gstack, you'll understand this is actually what's happening.

What I discovered is that—and this is more or less by accident—as I was writing half a million lines of code for recreating my startup that I created two other times previously but doing it in about five days, or during the course of several months creating Gstack, I realized that it's actually really useful to pull out specific personas of what is already in the latent space.

And so the most famous skill that a lot of people use—it's actually interestingly a distillation of what we already do at Y Combinator. At YC we have 15 partners, 16 partners at YC. When you have an idea and you're doing office hours with us, we're mainly asking questions about what's the problem, who's the customer, how do you know that, and then what are we building, right? And so that's what the Office Hour skill is. We basically took three or four months of transcripts across thousands of conversations, distilled that into something very, very potent, and then I had to distill that down by 90%, and then that's what is shipped in open source in /Office Hours in Gstack.

But as I went, it turns out there are lots of different things that I like to use to actually make it easier and far better to use the product that you can create with coding agents can be better if you're literally pulling out the latent space for a particular vibe and thing that you're trying to go for.

So, plan CEO review, for instance, my favorite thing about that is it asks the question, 'Okay, well, it has context, it knows what you're trying to build. What is the 10x version of that? What is the platonic ideal of that?' And so when I was a product manager at both Palantir and Microsoft and a founder for my startups, that was what I wanted to do. I wanted to figure out what is the perfect manifestation of the thing that we could build, and then what I'm building right now needs to be on a road map that is a straight line from where we want to go from where we are now.

And then the other thing that I discovered as we were doing this stuff is that you can boil the ocean. Who here remembers that term? Boil the ocean. If you go and work someplace, you're going to go into a meeting where people start saying things that are a little too scary, and then immediately people in that room are going to say, 'Whoa, whoa, whoa, let's not boil the ocean.' And my response to that, based on my experience with coding agents and what's happening right now, is actually let's boil the ocean.

The things that you can do—basically you sitting in front of one of these terminals can do the work of about 500 to 1,000 people. And if that's true, then all of the expectations that we currently have in society around what a founder can do, what a company can do, what a small team can do, what you can do sitting in front of a computer, they're actually a thousand X wrong, right? And actually what's funny is that's baked into the model weights. Who here has asked Claude Code before, like how long is this going to take? And it'll give you, 'Oh, it's going to take about 3 weeks to code all of this stuff.' And then you press approve on the plan, and then literally it's done in about an hour.

So all of us have experienced that. The models themselves have not caught up to this new reality that we can actually boil the ocean. So anyway, use Gstack. There's a lot of stuff in there. We have very little time, so I feel like I need to skip ahead. Gstack was basically my understanding of building open source and putting it out there. I'm still working on it.

But the new thing that I've been working on that everyone at YC has been completely immersed in is Open Claw and Hermes Agent. And they're actually teaching us brand new primitives on how to think about code, how to think about markdown, and how those things work together to do real work.

And so this is somewhat obvious, but I have to say it because anytime I would build an agentic system and it broke, it would every single time break because something was wrong about what I was trying to do. Like I was either trying to do deterministic work, like things that should be in code in my markdown skill, or I was trying to do latent stuff, like actually the things that my agent should be doing using the LLM in the code.

And a concrete example is, we spend a lot of time trying to curate the experience of people at YC events. Anyone—you can just use Claude. You don't even need Claude Code. You could use ChatGPT. Put in bios of eight people coming to your dinner party, and you can have it go and Google that person, run a dossier, and then figure out who should sit next to who. That's very easy to do in latent space.

But try to do that with an 800-person dinner party, or with the 6,000 people that are coming to Startup School. You can't do it. The model's not big enough. It hallucinates, it doesn't work. And so, what do you do? Well, that's the perfect example of—you need to make the latent space work with the deterministic space.

And so how do you actually do that? Here's a toy example: what is a skill? Who here has played with a skill or used a skill file? So, a skill file is actually—it sounds facile. If you go on Twitter and believe the haters, they're going to say, 'Haha, it's just a bunch of markdown files. Who cares, right?' But the big difference now with LLMs is you can actually do real work with this stuff.

The thing that keeps coming back over and over again is that you can do real investigations about it. And so basically, what is a skill? It's basically just a runbook. Even if you've ever thrown an event, and you need to throw that event over and over again, what do you do? You go into your notebook, and you just write down, 'Well, one, we need to secure the venue. Two, let's figure out who should come.' It's just this—any human being or agent should be able to look at it and say, 'Okay, after I read 1, 2, 3, 4, 5, 6, however many steps it is, maybe it's branching.' It could be very complicated, actually.

Do I know how to do that thing, right? This is a very simple concept, but the really cool thing is that you can actually make it call code. And that's what I find myself doing inside of Open Claw and Hermes all the time. And this is where it links to what you guys are doing as founders. And this is the pattern that we're seeing inside every YC founder or inside every YC startup now. We're not picking up the phone and doing it ourselves. Just like we're not opening VS Code and writing code ourselves. Every—Claude Code revolutionized how we write code, and we don't open the editor at all, right?

The same thing is happening with Open Claw and Hermes Agent. So all non-technical or process-oriented things in knowledge work are now—you can do it in Open Claw. You can have Twilio call someone. You can use Gemini Live to actually book a thing or buy a thing or 'here's my credit card.' All of these things—do you remember that Google demo where they stood up on one of their conferences and they're like so proud, 'Gemini can now call and get you an appointment' and then they never shipped that thing? You don't need to wait for them to ship that anymore because you can have that yourself. And that's the most empowering thing.

So code is code. The concrete example I have is who here uses Open Claw and it always for some reason thinks that you're in Greenwich in the UK. Like it's always—And so this is a perfect example of—I had to write code in TypeScript as context-now.mjs and I have tests for it. And then I have it built into my system so that I don't rely on the latent space to do it. It just tells me here's the time and then actually here's the things that are coming up. And if I don't do that, left to its own devices, the latent space will be like, 'Oh yeah, it's 3:00 a.m. Why are you still up?' And it's like, what are you talking about? It's the afternoon right now.

The next important thing that we discovered—anyone who has used Claude Code a lot has probably seen this error message at the top of Claude saying, 'Your Claude.md is 40,000 tokens or 40,000 lines or something like that.' And then you Google around, you're like, 'Okay, well, how do I fix that?' Well, how you fix it is actually a resolver. So, a resolver is actually really important because it's amazing how much you have to spend time getting this right.

Claude.md is a whole bunch of instructions—

Instructions on how to do things that you developed, like you got mad that Claude Code did this or that or wrote the changelog in a certain way. You say, hey, I don't want it like that. Don't do it like that anymore. Well, turning it into a proper resolver means that you take that instruction and it's like, anytime you have to write to the changelog, load changelog.md. And so, suddenly you don't need that in your context. The agent itself knows, oh, okay, here's this master directory of all the things I know how to do and I need to load the instruction only when I actually need it. It sounds so simple, but it's kind of obvious, but like this is actually the core of having a really great agent, actually. It's having a resolver. When I need to check signatures, I want it to actually go to my executive assistant skill, who is a particular person, like, well, I needed to look up in my brain repo how to do that. And I have a skill, a specific code path, and it's not a code path, it's like a markdown code path, right? It's a, I call it a skill pack. I have a skill pack specifically for that thing. I did it once, and then that's where, here's another primitive that I discovered that I find myself doing about 20 times a day when I'm using Open Claw or Hermes Agent. It's called skillify. So, it's, you know, you're sort of going up one level in abstraction. So, let's use one of these examples. You know, save this article. Well, I do that once. I look at the input, I look at the output, I get the agent to do exactly what I want. And then once I have it in a position where I like it, I actually tell it skillify. And then on the right, that's actually what the skill says. And, you know, this is a summarized version of it. I have an article on X about it if you want to see like all the full details. But long story short, you write the skill, you write the code, and then here's the part that is actually broken in Hermes Agent. I think they're about to fix this actually. But it's not enough to do it once. You actually need to test it. You have to, it's like kind of like if you work in a finance organization, like think about all the people, like 10 or 20% of people who work in some of these organizations just do compliance. And you're like, what are all these people doing? Actually, like in an agentic system, this is exactly the illustration of that. Like look at all these steps. Writing the skill and writing the code is only two out of the 10 steps. All of the rest of it is making sure that this messy system that is kind of more like a human system than perfect, beautiful, beam of light code can still work and do work that you want, right? Okay, so you did something in Open Claw, you made it work, then you say skillify. What does it actually do? Well, you have to write unit tests for the actual code. You have to write LLM evals for the skill file. Then you have to write an integration test. Then you have to make sure that there's a resolver trigger in agents.md. And then you have to test that. You need an LLM as judge eval to make sure that when that thing comes up, it's broad enough that it actually gets triggered. And then there's this other concept that you can look up in G Brain called check resolvable that is very important. You want it to be dry, don't repeat yourself, otherwise you end up with skills that do all the same thing. You need end-to-end smoke test and then, you know, ultimately you need a schema. You need to figure out where does this live in my memory and my repo. So, we're going really fast, but you know, that's why memory is actually really important. And so, my next project that is out now that I'm working on is called G Brain. It's actually a three-layer memory system built on top of what Karpathy already talked about with his knowledge wiki. So, I started with a knowledge wiki as well and then it started falling over because it just uses grep. And so, I had to add, you know, vector search, ARR fusion, backlinks. I added a graph database as a typed knowledge graph. I'm about to add an epistemology system so that we know that things are, they're like hunches or beliefs by specific people or world knowledge and I want to track when things sort of, you know, what's funny about maybe this is very specific to me. Like I'm super fascinated with the idea that people in this room are going to go on to, like your journey as a founder literally is that you have a hunch. You think that like the world needs X. Nobody believes that yet. But, you know, I want my knowledge system to be able to track like, oh, well, I heard so-and-so, this person in this room, this person in red shirt right here. He tweeted this and nobody else believed that yet, right? But, he's going to go and spend like a year, two years, five years proving it correct. And then, if my G Brain is actually working properly, it's going to spot that. It's going to be like, oh, actually like here's at Stanford there's this one person who believed X and then they manifested it. And so, I don't know. I for me like philosophically, I'm fascinated by knowledge systems like truly capturing what's going on. And that's sort of what we, you know, I think about this like I'm just building software for myself. Like this is the stuff that we have to think about and I don't know. If you spot in my voice like I'm excited about this because I'm building again and I'm building for myself. And then we're open-sourcing this stuff because we want all of you to actually be able to do it. I feel like I need to expand on like, you know, one of the things that G Brain does is like it's a very specific schema for my use case. But, you know, one of the last things I need to do before I go to V1, hopefully in the next couple weeks, is I actually need to make a fully dynamic ontology, which is a great buzzword that I've learned from Palantir back in the day. I mean, that's what we, you know, right now it's built, the schema is built for me, but there's no reason why it can't be built for you, whether you're a researcher, whether you're a journalist, whether you're a politician. Like, each person's going to have a different schema. We need to support all of those things. So, zooming out, I'm about to pass it over to Diana to take it all the way home. Like, I sort of gave you the primitives that we're learning literally like week by week. Like, I didn't even know about skillify until it flew out of my hands at like 3:00 a.m. using Open Claw. And then I put it on X and that went viral and I mean, I'm just learning as I go. I'm not an expert, you know. Sometimes it's like my favorite line from Alan Watts, who if you guys know Alan Watts, is he walks, he goes to a room like this, he used to give lectures and he would say, 'I am not a guru. I am just an entertainer.' So, you know, that's, I want to pass this over. I mean, we're talking about the agentic company. Diana's going to tell you a lot more about it. But, like, the concepts that I just talked about, like, one of the weirder things we realized is these actually map to the company. So, a skill is, you know, sort of a squishy human being who's an employee who has a capability. A resolver is the org chart. Like, who handles what? Like, how does it happen? Like, it's, you know, the filing rules, where it goes in the brain is the internal process. Where does the information live? Check resolvable is this thing that makes sure that the resolver works for like the set of things that you want to get done. And that's like audit and compliance. Like I, you know, when I was sitting in your seat, I had no idea why so many people in so many human organizations had to spend so much time on audit and compliance. But now at age 45 building a lot of agentic systems and looking at Skillify and how much time I spend just trying to make the things like freaking work, you know? I actually understand now. Like human systems are very messy and that's what check resolvable is. And in the end like, you know, the funniest thing is what a trigger eval is. Like you would think like, oh well, of course it's in the trigger, it's in the result, you know, in agents.md it should just work, right? But no, you even have to check that. Like that itself is its own latent space squishy operation that you have to check. And that's, you know, in an org, those are performance reviews. So with that, I want to hand this over to Diana to take us to the actual applied portion that will actually help you.

Yeah. Well, what's cool though is like once you get like the basics going, my favorite thing that I haven't released yet, but I will release is a cross-modal eval. So, you know, I'm about to add the skillify where you can actually have the frontier models of Opus, GPT-5.5, and DeepSeek-V4 all evaluate the inputs and the outputs and then rate it and then feed it back to the original sub-agent saying, you know, this is the rating and here's what you need to do for the next try. And then you actually iterate. You can meta prompt to get something that is 10 times better than the first version of what it is. I mean, this, what's weird is like these abstractions are basically stacking because that's what I learned from G Stack. A lot of YC founders said, 'Well, I like Claude Code, but that's like my ADHD CEO. And then Codex is my, you know, nearly non-verbal 200 IQ CTO. And I need both of them to do cross-modal analysis and then it ships with zero bugs.' So, these are all things that are like stacking. Like we're just discovering these things like week to week right now.

Can we meta prompt here for a second? Like you're sitting here listening to a lecture about this stuff, but the lecture is totally useless if you don't go and open your own Hermes Agent and open Claude and like load up your own G Brain and like actually use the, like there are 40 skills that you can test out and try inside G Brain. And some of it is like make your own. Like basically do stuff and then skillify your own stuff and then release it open source, too, and see what other people want, you know? Like that's, we're sort of like getting there together. And so the exhortation is like not only are we meta prompting the machines themselves, we need to meta prompt one another to be better and to be able to fuse with the machines in a new and more profound way every single day.

Not in the training set.

Yeah, we're at like the first pitch of the first inning on the revolution and you guys are the shock troops. And one other stat I want to give you from the last batches at YC is that in the past only the best top 1% of the companies grew 10% week over week. That was the metric that PG set. And in the past perhaps the batch of Airbnb only maybe Airbnb and another company hit it. But now what we're seeing things have dramatically changed where on average this is the growth of companies that within 3 months they basically 3x. Yeah, in the history of YC this has never happened before. So we get to live in this moment where like people in this room can create something that actually has a real impact and you can see it and you can tell because your customers are going to say, 'I can't believe this exists and thank you.' And they'll pay you and then every week 10% more people will be paying you.