The Exchanges

Q Incredible napkin math. Napkins are good. Napkins are good. So actually, what's a good napkin math that everyone here who wants to be a future founder should run tonight to potentially build something as consequential as the TPU?

A Yeah, I mean, uh, it's always hard to say. Um, I think, uh, Think about what problems you see and whatever it is you're thinking about, what, what bottlenecks you see, and are there very different ways of thinking of the solutions to some of those problems that would get you, you know, an order of magnitude or two orders of magnitude better, uh, performance or capability or whatever it is, um, you know, because sometimes if you just squint at a problem and you think about not necessarily being anchored on exactly how that problem is solved today, But how you would solve it from first principles, you can come up with really good ideas that are, you know, maybe not what other people are thinking about.

AI assessment note: “Think about what problems you see and whatever it is you're thinking about”

Q I guess selling data at the time, you know, large language models were not even, had not really come to the fore yet, Um, but self-driving cars were sort of coming up, and, and computer vision suddenly became, so that was sort of the first market, is that right?

A Yeah, so, the, the story here is that, like, I was, when I was at MIT, I did a bunch of projects to, like, train, train models of various forms, and these were, like, you know, by comparison today, they're, like, little toy models, and, um, and I remember to train a model, uh, I needed three things. I needed a, Uh, GCP account. Like, I need an account on some cloud service to get compute. I needed, um, the code to run to actually train the model, and I needed data. I needed a data set. And, uh, for two out of these three things, you could just press a button online and get them. And then for the last one, for data, there was, like, no effective way to get data for training these, training these models. Um, and so it felt, Incredibly obvious that this was going to be the future, that there was going to be a way to, um, you know, press a button, so to speak, and get data. And, uh, and it was very funny because in the years that followed, like in the first many years of scale, data was very unsexy still. Um, every time we would go out to fundraise, even though our numbers were great and we had great revenue, you know, VCs and investors would always be very skeptical. They'd be like, Oh, I don't know if this is a good business. Does it have longevity? Is this durable? Um, and, uh, it was really weird to me, but, you know, none of the investors had ever trained a model, so I guess t…

AI assessment note: “Yeah, so, the, the story here is that, like, I was, when I was at MIT”

Q later you, uh, came on to become president of YC and, um, you, you sort of brought exactly the same energy to, uh, a great many YC founders over the years. Um, did you have anything like, did anything jump out at you from that time around, you know, taking this raw energy of someone really, really smart and maybe a little undirected and then driving them more towards agency?

A Yeah, I, so first of all, I love startups. I think not everyone, they're not like for everyone, but I think startups are the coolest thing in all of business. Um, I think startups are really like the main thing that keeps the economy from becoming stagnant. Um, I think companies do just like drift towards suckiness, and startups will continue to be important forever. In fact, if, if we are right that AI is going to be such a big change, startups will be much more important to making sure that The power of this technology gets widely distributed throughout the economy and society, and it's not just concentrated in a few companies or models. So I, I think startups are this, like, unbelievably cool, fun, extremely painful and difficult, but wonderful thing, and I think a big part of the job of running YC is you are kind of the, like, unofficial flag bearer for the startup movement. Um, you know, like, there, there's lots of startups. There's lots of ways to do a startup. You obviously don't need to do YC, but it has always been a huge help to companies and a very powerful force. And so, starting with PG, uh, and then all of us, like, you know, we have to, like, make YC successful, but I think we really have to, like, fight for why startups are important and why people should consider startups and help put, like, relatively more power in the hands of founders and encourage More peo…

AI assessment note: “I love startups. I think not everyone, they're not like for everyone”

Q What are some practical things here? I mean, hard tech, for instance, I was hearing, ah, you know, at one of the breaks, someone was asking, like, should I go get my PhD? How important are credentials? Um, you know, how would you answer that, especially people who want to do these harder tech things?

A Well, as a general observation first, I think startups tend to win when, um, The sort of like technology landscape is moving very quickly when costs are coming down, when cycle times are short, and all of those things are happening right now. So if you look at when there have been like the great clusters of startups in the past, You know, there was the internet boom in whatever that was, like, 99, 98, 99, when new things became possible. Um, there was another version of, a mini version of people building on top of, basically, like, Facebook apps. There was then another big version when the iPhone app store launched. Um, But the great startups tend to cluster when the ecosystem shifts, and incumbents lose a lot of their advantage, and then also when you have this, like, cost and cycle time change. This moment feels very big for those things. It also has this other thing that you were talking about, which is a lot of the traditional things that were hard to get. Expertise, you know, the ability to go, like, hire excellent people that were That could do specific things you needed. That's really shifted, and in the last few months, um, I have seen a lot of people who just kind of grew up using AI the last few years, who are like, I can kind of automate an entire startup of agents and four of us, four people, and, you know, all this compute, and I think we're going to see much more …

AI assessment note: “a lot of the traditional things that were hard to get. Expertise... That's really shifted”

Q Who needs light? I mean, it's not so bright.

A What about these lanterns and what can you do with it? And, uh, and, you know, the steam engine is like, I don't know, maybe like 20 years away still, but, you know, electricity already exists. That to me was the sort of the, the, the, the fundamental, uh, Light behind it. And I would say, I think since then, OpenClaw was kind of a interesting sort of next step, which is, I think when we realized that, uh, you know, the reality is good AI products are agentic loops with tools. Uh, and we started doing that in our own product at Brex, but, but then on a personal side, I started spending a lot of time understanding, okay, what is at the frontier of, uh, using OpenClaw? And I think the insight was just, um, Yeah, like, markdowns can take you really far. Just, like, configuring and automating a lot of the things in your life. It's kind of funny. I remember I had this, this experience of, like, buying a movie ticket entirely in OpenClaw using, like, a BrexCard that was provisioned through an API, and, uh, and then I showed it to my team, and they were like, oh, but, like, you can go online and, like, book it in 10 seconds, and I'm like, that's not the point. You're missing, you're completely missing the point. Uh, but anyway, and then I went obviously very deep in this rabbit hole and, uh, started spending a lot of time thinking how to change the fabric of the company and the way we…

AI assessment note: “That to me was the sort of the, the, the, the fundamental, uh, Light behind it.”

Q And you have to hold, right? You kind of have, like, right? Like, one of the learnings is, yeah, go ahead.

A I think the more important thing that's going on deeper, which is that a whole bunch of important things are getting built, and you can, if you find them, you can fund them, and you can be part of them, and you can help create them, and create massive amounts of value, and the people that do that are going to get greatly rewarded. And I think that goes along with, Um, diligence. Like, you can't just, I think my perspective on this, and the way that I look at a lot of the space is, is that I think deeply about each of these pieces of technology, and I, and I approach it much more like investing into, into a startup, or investing into a project that I think is worthwhile and should happen, um, even if I lose all, all the money that I invest in it. And I think about the underlying value that's being created. Like what, what is this thing going to enable in two, five, 10 years from now? Um, you know, I think within the crypto space, you, you don't even need to think in 10 years.

AI assessment note: “I think the more important thing that's going on deeper”

Q apartment, right? We were, you know, cracking beers left and right. Like what a, what a rush. And yet, you know, the next day we didn't feel that way. It was just more, more revenue. And so, yeah, if you, if you take away all those fun little moments, um, where do you find the motivation to go further and to push harder? It's got to come from somewhere else.

A Yeah, that one's also fun because you know the lows are gonna suck, right? So it's actually, it hits you on both sides. Like, you know that there are gonna be moments you hate your life as a second time founder. But you also know, like, the first amazing hire would just feel like, yeah, that's one of the first hundred people I have to hire great, and then thousand people. You know, it's like none of these moments kind of stick as much anymore in your head, and that can be demoralizing. Um, I think one of the other tricky, um, advantages of first-time founders is that, like, Because they're going to have harder times typically raising money or hiring employees, they tend to have to innovate more. Like they, they tend to have to, they're constrained into building something good. Whereas like the lack of constraints so often lead people astray.

AI assessment note: “none of these moments kind of stick as much anymore in your head”

Q I guess next conclusion from this, which you have mentioned in the past that basically coding is solved, right? You have mentioned this. Um, I'm curious now that effectively everyone can write software, what separates the exceptional builders from the rest? What, what are the qualities now that everyone can ship code?

A I would give like one caveat. So coding is solved for the kind of coding that I do. It's not solved for everyone. You know, there's still code bases that are like super deep systems code bases where quad still struggles. There's distributed systems where quad still struggles. There's really kind of in the weeds UI verification, like something is off by pixel or something, but it's still not perfect at this. Like Opus five was a big leap in vision and computer use, but it's still not perfect. Um, but I, I'm actually curious for people here, maybe raise your hand if a hundred percent of your code is written using agents. You don't write any code by hand anymore. It's pretty good. Okay. How about more than 50%? Slightly less hands. Maybe about the same.

AI assessment note: “I would give like one caveat. So coding is solved for the kind of coding”

Q is the, this is the thing that people haven't really, we haven't gotten to as a field as much, but looking at how, how does the brain process information? Can you add new brain areas? Are there ways to understand how the brain is like, what, What is going on either to use this to build smarter machines or to think about how to treat things like depression or addiction.

A I'm taken by, ah, to what degree right now it's about sort of, um, taking someone who has a condition or a disease and then bringing them, like, sort of restoring them to, like, sort of capability, right? So I think that's playing out in AI right now as well, right? Like you had computers that had no ability to do, like, any sort of Pure cognition, or like, you know, and, uh, you know, no neurons, and then suddenly a bunch of neurons, and then AGI is sort of like what a human can do. It's sort of like a restoration of capability. And then, of course, there's like this other thing after that, which is, you know, uh, ASI, super intelligence. Do you ever think about what that might be down the road? You know, what is that for BCI?

AI assessment note: “Do you ever think about what that might be down the road?”

Q In particular, what are the, the additions that differentiate it from, you know, HN job posting, for example?

A Yeah, absolutely. So the way I think about hiring is like, it's fundamentally a matching problem, right? There's a great company for everyone, and for every company, there's a set of people who would be perfect for them. But it's really hard To make that match happen. And it's especially hard for startups. For large companies like a Microsoft, they hire people of all kinds. And so the, the matching just becomes easier because they're hiring for so many roles. If you apply to Microsoft, there's a good chance there's a role there that like could be a good fit for you. But with startups who are only hiring for a couple of roles that don't have brand names, you don't know who they are, what they do, if you'd be interested, what roles they're hiring for. And so the matching problem is particularly hard for startups. And that's why we've been working really hard on it. Um, Matt would be a great person to talk about some of the, the ways that we're trying to solve a mashing problem. It's still, it's still early days. It's a really hard problem, but I think we have some interesting ideas.

AI assessment note: “Matt would be a great person to talk about some of the, the ways”