Y Combinator CEO Shares How They Turn Wild Ideas Into Real Companies

I don't think people are prepared or even aware of what's about to happen right now. And it's like super, super good news for anyone who's running a business.

Sam, have you seen, I feel like there's a new meme that got birthed and I think Gary's involved with it now, which is the learn to cook loser. Have you seen this, Sam?

What is that? What is that?

Gary, can you explain this one?

Gary posted it. I don't even know like what it was a reply to, but It seems to be going viral. So it's a good meme.

I think somebody, Paul Graham was talking about cooking. He was like giving his kid advice and he's like, you know, when you have a girlfriend or something like that, like, you know, they love to come home to the smell of like a meal being cooked. And then someone immediately, you know, random Twitter troll is like, why are you raising like a feminine man? Like, you know, he should be masculine. He doesn't need to learn to cook. And then I think Paul's reply was learn to cook, comma, loser. And then all the builders like Amjad and others, basically like tech people were like, you know, this is actually good. General advice, like learn to cook as a high agency way of living. I think it's like the compressed idea of, you know, build things, actually make things, actually go do things in the world that are of value. Maybe I'm reading too much into it.

It's a more succinct version of make something people want.

Yeah, it's the punched up version. Learn to cook, loser.

Is Paul Graham like the definition of success? Basically built like one of the greatest companies of all time, YC. And then it started working well and then just bails to Europe where he's living in a, I don't actually know if this is true, but it appears he's living in the woods, writing and being with his family. And I'm like, is this guy, did he just, did he just pull the ultimate move and it worked out perfectly?

I think so. I think he really did.

What's, what's your version of that?

I mean, more or less like YC as a concept is like kind of a miracle. It's, you know, there are very, very few, um, places in the world where anything like YC ever happens. And then I just feel like I need to be like the watcher on the wall. It's like, okay, let's, uh, this thing is working. And then it's working for relatively mysterious reasons that are almost too obvious. I mean, it's at once a mystery and it's obvious. YC itself is like pretty earnest. Like it's just, make something people want, apply in this thing. We'll try to go through and sincerely find the things we think could be really, really big. And then we just give you a bunch of money and throw you in a room with all the other people who are, you know, sort of the top 1% of people trying to do that thing. And then somehow some magic happens where, you know, 5 to 10% of those companies become worth a billion dollars or more. And, um, you know, YC takes some equity, but not that much, you know, like 7%. Specifically for people who often wouldn't have access to that capital or that network, like by default anywhere, anywhere else. I mean, so it's kind of cool. It's just, uh, you know, 1 1 3 in a really unusual way. Um, it sort of resembles a college in some ways, you know, only instead of people paying tuition, like we pay them.

You know, when I think of good brands, a lot of people will just think of like, oh, funny or cool things, but when you think of like what BlackRock or Goldman, like some of these names, they're kind of like Harvard. Yeah. Or Harvard. They're like, they're a little ominous and like a little scary, but they're institutions. Do you ever purposely think, how do we become an institution? Or is this just all because of, you just said, let's have high standards. Like, was it on purpose or did it just happen because you had good values? How does that happen?

I mean, I think a lot of it comes back to, you know, earlier we were talking about Paul Graham and, um, The guy's just a very unusual, original thinker and very just clear communicator. And then I think the thing I learned from him in particular is that clear communication and, you know, just using basic logic, being able to explain, you know, sort of, you know, A to B to C, and then you zoom out and there's a really big idea in there. That's, you know, in essence what a really good pitch is. Um, you know, I think that that's where it came from. Like Paul was just very, you know, both frankly Paul and Jessica Livingston, his wife and co-founder, the two of them basically created this thing that feels different, didn't feel like another investor, felt very welcoming. I mean, this sort of, uh, viral tweet I think PG has about Jessica is like, you know, Jessica once. Cut a founder's fingernails before he went on stage at demo day or something, just because it's, it's like sort of, you know, at once very exacting, you know, we have very high standards, but then it's at once also very, um, sort of like almost family or like homey, just, um, welcoming. And, you know, sort of that mix of both very earnest and very formidable, like those two things combined created, I don't know, something that seems to create magic.

Isn't it wild, by the way? Like, just think about this. So one of my companies, I recently taken through like an M&A process and I talked to bankers and bankers. So we, we, you go first and interview bankers. You're like, why should I go with you? Why should I give you 4% of this transaction for 6 months of work here at the very end? And they're all about their differentiated process, how they do more, more, more, how they have the smartest people in the room and all the things they brag about. About what makes them successful. On paper, makes sense, but it's kind of the Midwit meme because if you look at YC, it's like, yeah, we invest in people before they even have an idea. A lot of the time they have no traction.

Often they have to change ideas.

We don't negotiate every deal. We just 7% for every single company. Same deal. That's it. The application. Yeah, it's like a one-page form you fill out online.

Oh, the interview. Yeah, we talk to you for 10 minutes Yeah, it's 10 minutes. Oh my God.

It's a 10-minute interview.

Yeah. And they're picking winners at a higher rate than anybody else. I think they're attracting winners at a higher rate than anybody, so it makes it easier to pick, but they're picking winners at a higher rate than every other investor, even though their process is simpler and less than what you would think you need in order to be the number one player in the space. And they basically, I think I've read the stat, it's pretty insane. You guys basically, every time you run a batch, you basically print $3 billion of market cap on average. I did some math and I really looked at the batch values. So, you know, like something like 5% of all the companies become unicorns, but every— and you just do that a couple of times a year. You're doing it twice a year. Now you're doing it 4 times a year. You're like, oh yeah, let's just print $3 billion of market cap 2 to 4 times a year. That is mind-blowing. Like, that is so insane.

Yeah. I mean, I think the craziest thing I'd ever heard was basically the Airbnb round where I think something like under $20,000 became $2 billion in the liquid IPO when Airbnb IPO'd. So, and you know, YC didn't even take the pro rata then. So it was pretty wild, honestly. Like, you know, easily could have returned $5 to $7 billion with a good portfolio management strategy. But, you know, I, that's the thing. I think, uh, simple things work actually.

It's, uh, if YC was a, a company, because I think it's been around like 18 years, you said. The last 18 years, you know, YC's got to be up there in like the top 10 Silicon Valley companies, things that got created that created so much value. Um, do you think about it that way, or is there— do you have a good lens?

That's what I try and, uh, get people to understand. Like, this isn't just, you know, working at a company. This is, uh, you know, I think we did the math on, um, how much money per employee YC in theory should drive. And, uh, I think we're something like $20-ish million per year per employee in terms of like, you know, sort of market cap value created in the world. And that's just the carry, like that's our take, like, you know, in terms of value created for our limited partners and, you know, sort of overall in, in sort of I think it's like close to $100 million per year per employee. So it's pretty wild.

Who are the LPs even? Is it like the pension fund types?

Oh yeah.

Just like a VC fund or is it—

People who invest in VC. I mean, the wild thing about YC is I think we're going to release a white paper on this pretty soon, but I haven't talked about this publicly before. Um, we looked at a 2-year period from 2018 to 2020 of people who invested in at least 3 YC companies per batch for those 2 years. The, uh, top, top decile predictably was very good. It was like 16x. Uh, the top quartile was also very good. It was 8x. Uh, I mean, these are outrageous returns for such a short period of time in venture period. The wild thing though is median was 5x and then the bottom quartile was 3.3x. So it's really crazy because in venture period, if you look at how VCs make money, it's all in the power law. There's only, you know, sort of the top 25% make any money at all. And then the top quartile generally for any given year of VC, uh, is only 3x. So it's pretty wild what's happening.

You've read, I think I read something, you've read 6,000 YC applications. So you've seen more of this than anybody else. You're the admissions officer. You've been in a lot of these, uh, interviews. You've then seen the companies grow. What do the extreme winners look like early on and do they look any different than the average company?

I mean, honestly, they're very similar. Like my two craziest wins, one was Brian Armstrong at Coinbase and he was a solo software engineer working at, you know, anti-fraud Airbnb. And, you know, he had never started a company before. He, you know, a lot of the things we worked on with him very early were like, sort of the basics that, you know, it's sort of hard to believe now, uh, on the other side of him having built Coinbase, it's kind of hard to believe that, you know, there was a time when Brian Armstrong didn't know how to, you know, do a product launch or, um, you know, even raise a single dime from investors. So I think it's basically been that same profile, like highly technical people who are just like sort of smart and earnest. And that's about it. Like, that's sort of mainly who we like to fund. Um, and then it's, I think, unusually really, really important today in 2024. I mean, large language models have basically opened everything up. I mean, almost anything that a knowledge worker can do, you know, anyone with 130 IQ or below, pretty sure that with Workflow and evals, you could write software now that could do what that you know, basically almost any knowledge worker could do and you could do it perfectly with like no management headaches. It's, I think, like a really profound moment for all entrepreneurs, period.

Hey, let's take a quick break to talk about AI. We all know AI, big deal. Um, you see demos all the time of people doing really cool things, but as a business owner, sometimes it's hard to figure out how do I actually use this? What do I actually do? I've been trying to use it across all my businesses. you know, things like making little prototype websites without needing to hire a coder or writing copy for our website. Or I give it a bunch of data and I ask it to analyze it for me. It's been kind of amazing, but the thing I always need is inspiration. I know the tool can do a lot, but it can almost do so much that I'm not really sure what I should actually be doing with it. And so that's why I think it's great that HubSpot has created a report where they surveyed 2,000 global marketing leaders and asked them Yeah, what's separating the high growth and low growth businesses and what strategies they're using with AI in their business. You can grab these strategies and apply them to your own business for free. The link is in the description below. Yeah, I think me and Sam have both had, uh, learn to code on our to-do list for like 12 years. Bought the, bought the, uh, you know, Python the Hard Way course on Udemy, did all that shit. I think we, now we get to skip the learn to. It's just code now. You don't have to do the learn to part.

It's like you don't need to actually write it. I mean, it helps to be able to write a little bit yourself, but yeah, I mean, the wild thing from here is with like the new developments in O1 and CodeGen, I actually think it'll be more important that you're able to speak to virtual developers and like instruct them on exactly what you want and all the different use cases. And like, we're right at that moment where it's going from like, you really have to code it. All yourself to, you can just sort of talk to this pretty smart entity that will, you know, do it mostly for you. Like it's not even about no-code anymore.

I was doing that the other day where I was talking to ChatGPT where I was like explaining like, here's the problems my business is having. Can you help me? Like, can you just like be a therapist with me and help me like come up with interesting solutions? And the answers were shockingly good, by the way. It was like, it was like pretty good at just like brainstorming with me and just being like a sparring partner. Is there a, you know, like I know that YC, a lot of it comes down to making people what people want and things like this. And to me, that's a, that's like a bit of an art where you're, you're more of an artist than a scientist. But is there a world where you could take your 6,000 applications and a bunch of other data that you have and kind of spreadsheet your way to like the likelihood that we build a company that, and I think there's magic to create these like multi-deca unicorns or whatever, but is there a way where you could like use your data and spreadsheet your way to like, well, here's the interesting opportunity. This is going to get to some type of traction and wealth creation with a high rate of certainty.

It's conceivable. I, you know, some of it is like, we think someone might do something like that and it might totally work actually. And then the trickiest thing is like, how do you actually get really smart people to come and do that thing? And, you know, I mean, capital as a service has been an, I, I think like Chamath talked about that and I thought it was very interesting.

What did he say?

It sounds conceivable. Uh, I mean, I think they experimented with this when he was still working on Social Capital where instead of, uh, having a 10-minute interview. And you know how it works at YC is we actually have 14 equal group partners, including myself. So we actually, you know, hand-read these applications and, you know, you're not talking to an AI when you do the 10-minute interview. It's literally just talking with us. And then once we fund you, we're just working with people day to day, you know, for those 10, 12 weeks and then onwards into the future of the company. Like we own, you know, 7% of the company. We're like a big seed investor. So we need to sort of be around and it's very like high touch and You get like an individual person. And then I think like the capitalist service idea, like, is really just, you know, connect your metrics, connect your Stripe account, your bank account, your QuickBooks. You know, I mean, it's actually pretty doable today. Like, it's conceivable that someone should connect all those things up and then an LLM can just, you know, rank whether something's real or not, do the diligence. And then if you like hit a certain bar, like, boom, like, here's money in your bank account.

Sam, did you ever see that when they did that for for growth stage companies. I think they called it the 8 Ball. It was like you, you connect all your data and then it'll, it'll not just tell you, they like, they could just take your raw data and they could put it into a format where they needed to analyze it. But then secondly, they benchmarked you. So they're like, oh wow, they have best in class retention for this category or they're below average. So don't invest.

Right.

But I thought the results were like interesting, but not like, like, yeah, where it was like, this is interesting. But do you remember when, um, ChatGPT started getting popular? I think Sean, 2 years ago and there was a guy on Twitter. Who had a really funny, like, name. It was like Greenpoint or something like that. And he told ChatGPT, like, make me a content website that can rank high on search and it can monetize on ads. And I think within like 3 months, he got it to like $20,000 in monthly revenue. Do you remember that story?

Oh yeah.

Like the normal way now is I prompt AI and I want AI to be an agent. What he did was the opposite. He's like, you tell me what to do. I'll be your human agent. You, but you have the good ideas. You prompt me. And I thought that was like kind of—

I bet it would work extra well now with, uh, O-1, like the new reasoning model.

Gary, can I ask you a question? You, um, today you probably think about like, you know, the future and billion-dollar ideas, but when you were 14, I read that you kind of had your own hustle there. You were cold calling companies in the Yellow Pages trying to get work, and you, you made enough money to kind of actually like help your family out, like move from an apartment to a house. Can you tell that story? I've never heard that before.

Yeah, of course. Um, I mean, let's see. I mean, my parents, we grew up You know, sometimes food insecure. My dad was a foreman in a machine shop and my mom was a home health aide and certified nursing assistant at convalescent homes. She worked two jobs and, uh, me and my brother lived in, you know, apartments in Fremont in the Bay Area, sort of in the shadow of Silicon Valley. And, uh, I guess one of the weirder things is like my dad's actually an engineer, but he just didn't have the EQ to really hold onto jobs. And he also struggled with alcoholism. So we just, he was always losing his job and so we were always just like never had money around. And the thing was like he really did invest in computers. Like we had our first computer was a, uh, you know, IBM PC XT. So, you know, we were sort of reality poor, but you know, rich in access to computers, which was like really, really powerful for me. And, um, I just realized Oh yeah, like I'm 14, but I, you know, know how to make webpages. And I started winning some awards for like web design as a kid in junior high. I started an underground newspaper with my, um, uh, junior high friends and, uh, we learned how to make a website for that underground newspaper and we got, and we called it an online zine. And, uh, the funniest thing was like we had, you know, thousands of people would find it off of Usenet newsgroups and things, and we'd write articles that, you know, uh, I don't think people knew that we were 14. Like, when, uh, when you're on the internet, one thing we learned is like nobody knows that you're a dog, I guess. I don't know. Like, people would just write letters to the editor like we were like a legitimate publication, and I just got sort of really addicted to that.

Like, like an early Vice.com?

Yeah, basically. Like, I think we wrote, uh, you know, my, my buddy wrote an editorial about, you know, how the three strikes law was bad for California. And, you know, we were like 12 years old, like What do we know about any of that stuff? But, um, you know, on the internet, nobody can tell. And, you know, I, I think we wrote it in pretty clear English. Like, people couldn't tell that we were that young. And then I, I feel like I got really addicted to the internet, and then I thought, well, maybe I can make money. And, uh, you know, back then there was this thing called the Yellow Pages that they'd throw on your doorstep, and that's how you would find businesses. And I thought, okay, the Yellow Pages is probably my link to making money somehow. Like, if they have a business and they have a little ad in the Yellow Pages, literally in the print section, like, you know, and this was not so early that like the internet had eaten Yellow Pages yet, but it was, uh, late enough that there was an internet section. So I just started cold calling, you know, seeing, hey, like, I know how to make web pages, will anyone hire me? And sure enough, I ended up getting a job and riding my bike to the local web design firm that made city websites for like City of Pleasanton and Fremont and San Jose. And, uh, you know, basically they paid me, you know, $10, uh, $7 and $10 an hour to work on, I mean, I learned how to do graphic design. I learned HTML.

What is that? Uh, what's that phone script?

Oh God.

Yeah.

First you have to deepen your voice, right? You can't be, you can't be 14 on the call.

That's right. And it's like, oh, I'm, I'm a web, I'm an award-winning web designer. And you know, but I'm 14. So did you say that? Yeah, I think so. I mean, um, it's funny. I mean, I remember my first boss, he was, this was sort of his like side hustle.

So by the way, is award-winning designer in your LinkedIn now?

No, I don't know if it is. I should add it again.

We'll add it back.

Yeah.

That's a miss.

It's funny. I mean, I remember it was like this company called InfoLane. It was, you know, they had a one, like basically I think it had a conference room and you know, sort of an engineer den and a lot, and like a little reception desk. And that was the whole office. And, um, I think it was one really technical, like sort of neckbeard Unix guy. And then, uh, the funny thing was the CEO was also a bank, a regional bank manager, I think, at Fremont Bank. And so, you know, the tech guy had walked into the local Fremont Bank and, you know, the internet was happening and this bank manager was like, I'm gonna give, I'm gonna invest myself, like I'm gonna invest my own money. And he became like the co-founder on this, you know, web design firm in, uh, 1993, '94. So it was pretty wild to, to think about. I mean, that's just how you started business and that was what business was like. You didn't go on the internet and apply on a form or something. You literally walked into a bank and banks were supposed to loan you money to start companies, which, you know, I think to some extent still happens, but You know, it's just funny to think like, uh, how that's evolved since.

And at some point, did you just make enough where you were like, mom, dad, here, like I can help? Or what happened?

Yeah, I mean, we, I always lived in apartments growing up and I had a little brother who was 8 years younger and, you know, I wanted us to live in, I wanted the American dream. And so, um, you know, I still remember like, uh, helping cook the turkey at Thanksgiving and, you know, the new house that that, you know, our family bought. And, you know, I helped my parents with the down payment, and my parents still live in that home in Fremont. So I'm helping them remodel it right now.

That's amazing. By the way, I looked up your old blog. I think back in 2001, you wrote the following, which I think is pretty accurate. You said, you know, blogs are all over the place. This is back in 2001, so over 20 years ago. Pretty soon we'll all just be writing a blog. We won't really talk to friends or have to catch up anymore. Human interaction will just boil down to clicking on someone's name on your favorite website, and that'll be it. No more tedious how do you do over an iced matcha frappuccino at some pretentious cafe. Just read my blog if you want to know what's up. Oh my God, you basically predicted social media. It wasn't exactly blogging, but that's Instagram stories.

That's Facebook.

That's the news feed. Exactly. I think you nailed it, for better and for worse.

That's funny, man. I wish, um, The one thing I wish I could do is like go back in time and tell myself, don't sell your time. Like, don't work for a consulting firm or like, you know, get paid $10 an hour. Like, go make products.

But did you, did you ever do that? I thought you were pretty young when you first started getting involved with YC.

Oh, uh, I was 27, so I was much later, honestly.

I mean, you looked very young in the photos.

You actually didn't need to come from the future because I think I've heard a story that you were working somewhere, Microsoft or something like that.

Oh yeah.

And Peter Thiel was like, Peter Thiel was like, hey, quit your job. In fact, I'll pay you your whole salary to quit your job and come co-found this new company. This little company we got called Palantir.

Yeah. It's like a $95 billion market cap company today.

And you turned them down. Could you, uh, I mean, what the hell were you thinking? Could you just embarrass yourself here?

Of course. I mean, I, you know, I went to Stanford, computer engineering, uh, I mean, changed my life, obviously. Fraternity brothers of mine, Joe Lonsdale and Stephen Cohen. Stephen and I went to high school together, grew up together, tried to start a bunch of companies together. You know, I was a year out of school, it was 2004, and I thought like, oh, look at me, I have health insurance, I have a real job. And then my old fraternity brothers who were a couple years behind me at Stanford, like, I guess they couldn't get a real job, so they're going to, you know, go start a company. I don't know. It's really funny to think about in retrospect. But, you know, one of the things that was always really smart about the Peter Thiel world is when you start something, you always just need the smartest people around you. And I was lucky enough that I made, you know, Joe and Stefan's list when that happened. And, you know, Peter was just running the playbook that worked for him when he sold PayPal, which is, And this is something that people at Palantir do all the time now. It's like when you start a thing, you, you know, take an 8.5 by 11 sheet of paper and you write down all the smartest people, you know, and then you go take them out to dinner and lunch and just, you know, sit on their doorstep until they quit their jobs to come join you. And, uh, you know, I guess it was pretty wild. I mean, Peter was not a billionaire yet. I think like the funniest thing about PayPal and the PayPal Mafia is that like they basically fought a land war in Asia. Against eBay, you know, I mean, when you don't have, when you have to work on other people's distribution and other people's platforms, you just end up becoming like the most vicious, hardcore, you know, builder fighter types to just like gut it out to build a business on someone else's platform. And then, uh, the exit for PayPal was just not big enough that it made people, it made people like able to take lots of risk. but not so much that they had lost all energy and will to fight.

Well, and there was a bunch, and there was a bunch of them too. So that got divvied up a bunch.

Yeah. It's wild.

How much was Peter Thiel worth, do you think, at that conversation? Because like, let's say you made $80,000. That, that's like a big deal, right? To be out to dinner with someone and be like, I'll give you 80 grand right now just to quit.

Yeah, totally. To get a co-founder or co-founding engineer, like I, I, I, some of it is like, I could see myself doing that today. I mean, I think Peter was probably worth like at least a couple hundred or at least $100 million. At that moment.

More than enough to, to be able to, to say something like outlandish like that.

Yeah. I mean, I think it's a fair trade today. Like I would definitely trade like my, you know, my gold Rolex for a first co-founding engineer of a company that I think could be worth $100 billion in the future. That is a fair trade.

Did he make a compelling pitch like about the company or was he just like, I don't know, this is just like a bet. Like, was he, did he see the future in that, that sense of like what it became?

Yeah, absolutely. I mean, I think the thing I really learned from that moment, I mean, I was 22. I didn't know how the world worked at all. I mean, we had dinner at his French restaurant Frison. I think that's one of the things you do when you, you know, make it to Send to Millionaires. You do something stupid like open a restaurant. And I mean, the restaurant was terrible. Like the soup was really bad.

And I mean, he lost like millions of dollars on that restaurant, right?

I think so. I mean, today it's this really nice steakhouse, but Yeah, it was bad. Um, and then, you know what I should have done? I mean, I was so lacking in social graces, you know, they did the whole pitch and then I think I said no, like before the soup came out even. And so we were just like awkwardly sitting there for the rest of the dinner. You know, and he had actually re— he was like, here, like, here's a check, like $70,000. You know, how much a year do you make at Microsoft? Here's $70,000. Cash this check. I didn't really understand it in the moment. And then I think in retrospect, like sort of the, the number one thing that I didn't understand that I do now sort of realize, like I have to tell, you know, tell everyone this is how it works is, um, you know, I didn't join at the time because I thought that, um, the world was sort of told to you or like, you know, sort of narrated to you, I guess. You know, I, I thought, oh, the Wall Street Journal isn't writing about enterprise software. This isn't a hot space. And so, you know, that, that was like fundamentally the backward way of thinking. Like, the world doesn't like happen to you if you are actually a like prime mover in the world, like Peter Thiel or my friends. Um, you go out and you go into the world and you discover something about the world. Like, you actually talk to people, like you actually talk to your users and, uh, you I sort of gather primary first source information on how the world works. And what they had done was realize that the world's biggest companies and the world's biggest governments had no access to software engineers or, you know, big data. Back then, people said like big data or data mining was like, you know, today we say AI, but either way, these were just buzzwords for concepts that are like actually pretty real, like I think they, you know, went into the security apparatus of like the three-letter agencies in DC and realized like all the people they talked to, you know, they were working, even CIA, NSA, like the, you know, people who we expect to have the world's best technology, they really don't.

And Peter knew that because he somehow, the CIA, NSA, whatever, when PayPal was running, they're like, uh, you guys want to have a conversation? Or like, you know, how does like a, like a Silicon Valley kind of, uh, outcasts get a CIA tip?

Yeah. I mean, my inter— I, I don't know specifically. I mean, how they described it to me, I think was like they learned a lot about how the world works through the anti-fraud part of PayPal. And the core idea was that, um, they could find these fraud rings just using a graph database. Just literally, you know, people make a bunch of fake identities. They send the money from one fake identity to another. And, uh, if you're just using SQL or a row-based method, you just couldn't figure it out. But the second you like graphed it out, you would find this ring. And you know, uh, what they found was that, you know, the, the three-letter agencies didn't have access to that technology. So, you know, what I needed to do was instead of taking my worldview from the media or, you know, today from social media or from, you know, what reporters were sort of distilling for me. What I needed to do was listen to people who had actually lived direct experience in doing something unusual and strange and life-changing for people.

That's simple, but that's still really hard to like—

Oh yeah. Well, it was smacking me in the face and I, I didn't—

You had an L6 promotion waiting on you, right?

At Microsoft.

Yeah.

I was trying to make it to level 60. Yeah. Level 60 is like hardly L6, it's like level 3 or something. It's brutal.

And by the way, Palantir, I've got friends that are there and I've learned a little bit about it reading. Like it did not— Peter Thiel, I think it's easy for me to say Peter Thiel back then seemed like an obvious winner. But the Palantir idea didn't— like it seemed, it seemed like it didn't work for like 2 years or something like that. Like it was definitely, it seemed like it had this normal startup trajectory where it's like, it's not working. Maybe it's working. It's definitely not working. All right. Now it actually has legs, you know, the YC Trafisaro thing. Like it seemed like it had those normal, normal like issues, except for like maybe 5 years.

Yeah, it took many, many years before it started driving revenue and, you know, became real. Yeah, the trickiest thing, like YC companies always come to me and ask like, what could I learn from the Palantir experience? And I'm like, are you, is your name Peter Thiel? And are you willing to put, you know, tens of millions of dollars of your own money into it before and sort of wait and be incredibly patient and ignore all the signs that you're, you know, weren't getting revenue and not, you know, actually working for so long to the point that, you know, suddenly it started absolutely working and working like 10 times better than you thought it would. Like, that's just not a thing that people are really capable of doing on certainly their first million dollars or their first, you know, Palantir is sort of like the, the most amazing version of like the second or third startup that people do. Like your first startup, you gotta just gut it out sometimes. It's like, how do you like, get to your first million dollars, right?

Well, that is the name of the podcast.

Indeed.

And so I'm curious, what's your advice or take on that? Like, how did you do it? How did you think about it? And knowing what you know now, how would you go make your first million?

Yeah, I mean, what's funny is in the end, like, Palantir ended up being huge, but it took— it didn't really IPO until many, many years later. It's like, I mean, I designed the logo in 2000. I joined as employee number 10. We would do these ridiculous things like go to Stanford and give out pizza. And then we were like 10 people, 15 people working there, but the pizza box would have our Palantir logo that I designed and like, join the next Google. And people would like eat the pizza and they'd like give us dirty looks like, how dare you? Like, what the heck is wrong with you guys? And then, um, let's see, my first win, I mean, the Palantir shares did turn out to be something, but it did not seem like it was going to for a really long time. I think my first actual million dollars was selling a few million dollars worth of Twitter stock at IPO. And so it was actually really late, like 20— I mean, for me, I started working in tech like when I was 14, so like 1996, '95. And then, I mean, I think it, yeah, it literally took like 18 years of like working as an engineer and not having money and having like $50,000 credit card debt coming outta college. And I think that might've even been one of the reasons why I didn't quit to join Palantir. I was like, it, it felt, you know, in 2004 I still had like $30,000, $40,000 in credit card debt and like maybe $50,000 in student loans. And I just like, I was like, well, like $70,000 is a lot, but you know, how am I gonna actually, you know, I don't know if this job will even exist in 6 or 9 months. And like, I could be a lifer at Microsoft. And of course in retrospect, Like being a lifer in Microsoft would be sort of like a life of torture at some level. So yeah, my first million was actually Twitter stock. Uh, so I started a company in 2008 called Posterous. It was a blog platform.

I used to use it. I love Posterous.

Yeah, it was, uh, dead simple blogs by email. Um, you just emailed post@posterous.com. You didn't have to log in or like learn how to use software. And then we just reply back with like, here's your blog, like here's your website. And you know, you could send it to your friends and You could, it was sort of like an email list too, so you could do it with your family. A lot of people used it to like share photos with their, uh, family. It grew 10x year on year, 2 years in a row until, uh, Instagram came out. I remember when Instagram came out, Posterous was actually one of the little checkmarks. Uh, it's lost to the sands of time. I think it was like, you know, Posterous, Tumblr, um, Foursquare, Twitter, and I remember all that sharing page. And I remember Chris Sacca, one of our investors, emailed us about it and was like, hey, what do you think about Instagram? And before I could reply, my co-founder replied, it sucks. I was like, and I was like, and we got a one-line reply from Sacca saying, I'm very disappointed in you guys. And then I just was like, you know, sleep deprived, didn't think about it again. But like, I think about that all the time now. 'Cause you know, when you're working on a startup, like, You know, we didn't actually even totally understand why we were growing that fast. Um, in retrospect, it was actually because the iPhone was new and people were taking lots of photos and there were no apps that made it easy to upload. And then Instagram was sort of the first free app that created a network that also even had a little bit of utility with their filters, and it was all free. And, uh, basically, you know, we got run over by the eventual winner. And I, you know, I think that's one of the big dangers for founders, period, is when we launched in 2008, you know, Michael Arrington at TechCrunch wrote about it and he framed us off the bat. He was like, Posterous is, you know, way easier to use than Tumblr. And so we sort of made a mistake and took like that competition and then made it our identity. And it's something that— advice that I have to give to founders all the time now, which is like, we weren't competing against Tumblr. We were competing against how people were getting photos off of their phone actually. But you know, when you're not careful, you sort of allow media or other people to sort of frame your reality and then you just play the wrong game. Like Tumblr didn't even wanna be Tumblr. That ended up having zero enterprise value. I think that's also one of the cool things about Silicon Valley is like you can even sort of screw it up. You know, it, if I really, uh, introspect on What I wish happened, um, you know, we had about a million users who were like super dedicated to Posterous. We easily could have, you know, sort of charged $5 or $10 a month for it. We would, you know, we were a team of like 12 people. We easily would've been profitable. And then, uh, I, I remember because we were hanging out with our friends at Weebly, like, uh, David Rusenko and Dan Veltree, who were amazing. I mean, they, they were, they only ever raised, uh, you know, a tiny amount of money out of Demo Day and never raised a Series A.

And what did they eventually sell for, like $300, $400 million?

Yeah, to Square, like sort of, I think, pre-IPO. And then Square, of course, like went on its crazy run. So these guys are like, made, you know, 100 times more money than me and my co-founder Impostor is like, you know, rightly so. They did it not through like continuing to raise venture dollars, but by, um, being profitable, cash flowing, by compounding, by focusing on their users and like playing their own game instead of playing the game that was like, can be sort of chosen for you by either your investors or social media or by reporters and tech, tech press. Like, you know, just play your own game, like make up your own mind. So, you know, if I did it again, like we easily probably could have done that. And then instead, I think we ended up pivoting out of, um, you know, that posturous idea and, you know, we were just wanting to continue to chase this idea that we could be the free social network. And I distinctly remember meeting Peter Fenton at Benchmark and he passed on our Series A because we didn't have a good answer for this. He asked, you know, are you a platform or are you a network? And, uh, we said both. 'Cause you know, that's what founders who like don't have strong opinions end up saying.

I think like it's a multiple choice test. You just circled A and B. Yeah. Can't be wrong that way.

But I don't know, this is all, you know, 20/20 hindsight. I mean, we use, I use these stories all the time to try to help founders get on the right path to real enterprise value all the time now. All right. So, you know, it's not just make something people want and raise a Series A and B and like, look like a successful startup. It's actually. Makes something people want and that's the end in itself. And if you do that, like you're going to make money and, you know, hopefully you can compound that and like, you know, create real enterprise value for yourself. And that's sort of the game.

We want to ask you a bunch of questions about different stories because you probably have a lot of stories. But what's interesting about that story is, um, you left Palantir as a very— I don't know if it felt it that way then, but now it's a very serious company. Where like they prevent terrorist attacks at times. So like, it's a very serious product. Uh, what you were working on after that, you could like, someone could be like, well, that's just like a blogging platform, or that's just this, like, that's, that's quote less serious. Did you ever, like, I mean, I guess I'm self-projecting because I feel this way sometimes where I'm like, I'm doing something that's not serious. But you kind of have done both and you've done, now you're back to a relatively serious thing. It's a YC's a multi-billion dollar company. Um, how do you balance like, like working on something that seems fun and cool versus like a serious big thing?

I mean, I think that goes back to what we were saying a little bit earlier, which is just like, be careful what frame you accept. It's all made up, but, uh, you get to make it up. And then, um, you know, for Posterous and blogging, like they said the same thing about Facebook, and I remember spending time with Boz and you know, a bunch of the, you know, our friends who were early Facebookers, they never accepted the frame that Facebook was not a serious and important thing. You know, they, their narrative was very clear. Like they had a very strong cult and the cult was just clearly once every few decades, I think very explicitly, Zuck would say this, like once every few decades, like everything would change. Media would totally change. And like, they were right. Like this social layer came up and, um, you know, this is sort of the reason why I feel like if you're running a business today, you kind of have to be a creator because if you're not a creator, you gotta buy eyeballs somehow. And, uh, guess who's got the monopoly on eyeballs? You know, Larry and Sergey and Mark Zuckerberg. You gotta go pay, you, you, you gotta pay the toll to, you know, go across the bridge. Like, you know, the second price auction mechanism in the ad networks for like all incremental attention is so valuable that these are like the most durable, powerful network effects businesses that like cannot be disrupted.

So wait, are you telling all YC founders to go and build an audience online?

Oh, if you're doing consumer, you basically have to, right?

What about for—

sometimes even if you're dev tools, like you actually—

but yeah, wouldn't you say the same thing if you're a B2B? Like, isn't it even more valuable because you need a smaller, more targeted audience?

Oh yeah, that'd be great. It's harder to target.

That's it. That's a pretty interesting thing. You didn't say this. I'm putting words in your mouth, but if you are saying to all the—

Oh yeah.

Yeah.

We have Mark Pincus coming to speak at our mini conference later today. And, you know, we're going to talk about it. It's like, how do you actually get incremental customers? And distribution really matters for consumer businesses, but that's also why, you know, we've had a dearth in really breakout consumer businesses for 10 years. Um, you know, I, I think that this will change. I mean, with AI, AI does change things. I don't think it's broken the distribution advantage. You know, I think that the big tech companies have a total stranglehold on the incremental user and you have to pay them still. Uh, I mean, it's just purely market dynamic, right? Like, you know, we're selling Widget X or Game X or Dating Site X or whatever it is, but you know, our, you know, buddy from college or, you know, or someone from halfway around the world in Eastern Europe. Can like take what we've done and like do the exact same thing, or go to our supplier in China and like get the same dropshipping relationship. And so at that point, like the second price auction means like, well, I want that customer. No, I want that customer. No, I want that customer. Guess what? It's an auction. Like every incremental dollar that is gross margin, Google or Facebook will just extract because it's a perfect marketplace. And that's just the world we live in right now. And then You know, I think that the purest form of alpha that exists still is saying interesting, valuable things that, um, you know, and the YouTube algorithm and the X algorithm and, you know, um, to a less extent like Instagram. I mean, I think TikTok is actually pretty wild. Like they're just giving away distribution over there, especially for brand new creators. Um, I think Meta is probably really dropping the ball on how they're thinking about creators. They just like, I don't know what's going on over there. Seems like they only want to satisfy the people who are already big, which, you know, is not that helpful for people who are trying to do new things. Um, but yeah, that's, that's sort of the regime we live in. Um, you know, for B2B, what we see at YC is that, you know, probably 70 or 80% of companies are B2B. And, uh, our most well-attended mini conference every batch is actually the sales mini conference. Um, you know, uh, Pete Kazanjee, uh, Founder Sales, like that book, we give a copy of that to like literally every founder. A lot of it is about a mindset shift. Like most people have never been rejected 95 out of 100 times. Like it's just so like soul crushing to get no that many times in your life. But it's also, you know, for 80, 90% of businesses today, you better get good.

If your website conversion rate's 5%, you're doing phenomenal. You go out and knock on 100 doors and get 95 nos, you feel like you're the biggest biggest failure in the world.

And it's the same comparison because you got 5, you have 5 enterprise contracts paying you $10,000 to $100,000 a year. Like, you're the biggest winner in the world. Like, that's how this stuff works.

We gotta ask you about some of the people that you've been able to meet. I'll tell you what I want, and I hope that you give it to us. I don't know if you will or you won't, but like, so me and Sam kind of are in what I'll call like the way that most people live their life, which is you wake up every day, you got your project, your business. It doesn't matter if you're doing, you know, $10,000 a month or $100,000 or $10 million a month. Doesn't really matter. You wake up every day and then we all look for kind of like this entertainment inspiration thing and we get a lot of it from Twitter, from podcasts. And one of the best things is when you hear about the way that some of the people you admire operate or the way that they, they handle things or the way that they dealt with a situation or something about their personality. Of course, it's sort of, you know, a caricature. It's not a perfect picture and, you know, whatever. They're not perfect people, but it is badass to hear those stories. And we hear them about Elon, but you run in a really amazing circle where, you know, Paul Graham, Sam Altman, like all of the YC network are people you've been around. You've been around them before they were, you know, big. And, and, and since I would love if you could tell us a couple of like either personality traits or stories that stand out for some of the people. So I'm curious. It's a long-winded way of saying, can you tell me some cool stories about some of these characters that you've run into?

Yeah, absolutely. The funniest story I remember from working with, uh, Paul Graham in the office, I mean, he gave me my first shot as an investor and as an advisor to startups. I was burnt out from Imposterous. We, my co-founder and I had like a falling out and, um, he wanted to become Google Groups and I was like, dude, I don't even think that Google wants to be running Google Groups. Like, why? We should just charge money. But, um, yeah, Paul, like, I think really did, like, Paul and Jessica basically, like, did all the admissions, worked with all the companies. It was like 2011, you know, even the whole YC campus was not a campus. It was like a warehouse with some carpet and some, like, custom, like, very cheap, I mean, it was just tables and benches. The benches were so rickety that, like, if you sat on one side, like, and the other person on the other side, like, you know, stood up, like you'd fall over. Like, there's just like a bunch of funny things about YC early. It was like this tiny thing. Um, and to paint a picture, like, people still, when I said I was gonna go work at YC, uh, as a designer in residence, my friends in venture or startups would say, oh, that's nice. It sounded like I was going to volunteer at like a high school. Camp or something. It was like a high school basketball camp or something. I was like, oh, that's, oh, that's nice. That's so, you know, I hope you enjoy that. Like, you're doing community service. Good for you.

Good for you.

Exactly. And, um, you know, Dropbox had happened as a, you know, and it had become a billion-dollar company, but, uh, Airbnb I think was just about to become a billion-dollar company. And, uh, you know, I, I, I think this is lost to the sands of time, but I remember Sarah Lacy had created a had written a book called like One Year Lucky, Two Year Good. So that, you know, 2011 I think was the one year lucky, two year good moment for YC where people realized, oh, this isn't like some fringe thing. It's actually starting to churn out like basically the most dominant startups that exist. It's like a very concentrated form of Silicon Valley. And, uh, about that time, um, yeah, we had, there was literally one person who did all the books, all the finances, all the audits, all the CFO stuff. And she needed a copy machine, so she got one of those, like, sort of waist-height copy machines that, you know, looked corporate. And it had, like, these little stickers on it. And then I remember Paul Graham came in. He's like, "What is this? Why is this here?" And it was like sort of this totem of corporateness. He's like, "I remember seeing these at Yahoo and I hated them." Like, he started, like, scraping off the sticker. That had like the phone number of the little company that, uh, would maintain the, the copy machine. And, uh, I guess it just jumps out at me as like this very interesting quirk of Paul that was like absolutely right. And, you know, I sort of stay up at night thinking about like, how do I make sure that YC never feels corporate? PG, like PG would show up in like shorts and, uh, Birkenstocks all, you know, all day. I think he, He wore khaki pants to my wedding, so my mother— my, uh, my sister-in-law was like, excuse me, sir, this is a private wedding. And I'm like, no, no, no, that's my boss. Um, so I mean, there's this— I think what I learned from Paul was that, um, I think there is an insidious nature to like formality and, um, corporateness and just like having to wear a suit or the convention, right? Like, I think there's incredible value in being very, very unconventional. Whenever there is prestige and/or convention, you should be wary of that thing because, you know, there's gotta be some other weird soul-sucking things that are associated with that, that, you know, no one by default would sort of choose to have, you know, a dress code or like, what, you know, what are some other examples of this? Like, you know, basically you should beware of about your own formality and, um, try to fight against it. It's just like, try to be super matter of fact. Um, and I think that that's, that's one of the things that just makes companies bad. I wonder, like, as people who run your own businesses, do you ever think about that? Like, you go into a meeting and you just think like, was that a meeting that if I were not me running this place, like, would I want to be in that meeting?

Well, it's the same thing as what you said about the framing thing. You, you behave like you think you're supposed to behave versus what Uh, kind of makes sense. Um, you start playing business a little bit. It's sort of like, you know, when you start a company, a lot of people play business by getting like a business card before they get a customer. And then you kind of snap out of it. You're like, no dude, I gotta make the damn thing. You start making the damn thing and you start getting some customers and then you kind of go back to playing business again.

Yeah.

Um, which, which is definitely part of the Silicon Valley thing, you know, like OKRs, measuring things quarterly. Like there is a bunch of like lameness that. I myself fall victim to, and you have to like pull yourself outta that trap. But, but some of it is like useful, I guess. If you're Airbnb, you, you have to, if you have 6,000 employees, you, you do slow down. But it's hard to know what's truth versus what, like, or what you should do versus what you have to do type of thing.

I will say that I think, um, it's possible for people to do, uh, insane business now with like 2 pizza teams. Like I think that we, you know, probably in your community, Or maybe you guys yourselves will end up making these like $100 million to like billion dollar a year businesses that are totally empowered by large language models that like do not need more than 20 people working at them. I think that this is literally the most exciting news that, you know, we're pretty sure is going to happen the next few years, especially right now. And like, that is the one thing that. These giant companies that have like thousands of people, like it is completely like corporate America is completely unprepared for this moment. Like they're just going to get run over by a thousand startups that are way more agile that, you know, aren't completely drowning in convention. They can just do the right thing for the customer and the cost basis will come down a ton. You know, the, the trickiest thing is like, I'm not sure if it's inflationary or deflationary, like what's going to happen here, right? Like prices in theory should come down. Things should get a lot more, uh, competitive. I mean, on the flip side, like, the hope is that all of our, you know, products and services get, um, a lot better, cheaper, faster, real fast. I mean, this is the moment. Like, literally, we talk to people who are eng managers and, like, CTOs and, like, even engineering teams. I, you know, I think that it's almost generational. Like, we get— it's, like, in the water for us right now because we spend all of our time with, like, 22, 25, 28-year-olds who were like the me from, you know, 15 years ago. Like, we're just helping the next generation right now. And they were like born on the internet. They were born with large language models, you know? Um, and then there's sort of, uh, you know, frankly, I'm 43, like my generation of C-levels, like, I don't, like, I don't think people are prepared or even aware of what's about to happen right now. And it's like super, super good news for Anyone who's running a business who's super agile, cuz you don't have to be doomed to calling in the business process automation call center. You don't have to like hire the team in the Philippines anymore. If anything, like that's one of the more direct advice we give to, uh, people looking for startup ideas in the batch right now. If people are pivoting, we're like, okay, well you're selling to, let's say like accounting firms. Um, you know, but you're having trouble with enterprise sales. What if you targeted people who already spend, you know, hundreds of thousands of dollars a year on a call center in the Philippines or Eastern Europe or anywhere else? Like, why don't you just reach out to people like that and just replace what they rely on, you know, there with large language models that have great evals and great workflow? Like, that is just, you know, I think we will have many companies driving tens to hundreds of millions of dollars, like, literally in the next like 2 years. It's just like hundreds of them, thousands of them are going to pop up right now just doing that.

That's so exciting.

What companies, um, are you super excited about that we should watch out for, we should go check out? Because you live in the future, you're, you're further in the future than we are.

Obviously, like, uh, the tooling for this is important. Like, there's going to be a lot around, um, helping people build the stuff. Um, evals in particular, like doing it test-driven is important. You know, um, the danger for most founders right now is it's very easy to make demo ware that you can use to raise money. And then you're tempted to sort of, you know, quote unquote raw dog your prompts. Like you will just write code that has some prompts and you won't write any tests. And then, uh, the funny thing that we're seeing right now that I think is super true is Again, like going back to what we were saying earlier, like it's important to build your worldview on how the world works directly from customers and their customer data. So the sort of rinse and repeat thing that's like maybe 80% of YC, what YC companies are doing right now that I think anyone could do is go and find real people who like are running businesses and they're spending hundreds of thousands of dollars a year on giant teams of people doing like rote repeating knowledge work, and then get access to their data, get access to the flow of their work, watch how they do it, and then literally write test cases using, and if there's hallucination or, uh, the LLMs aren't actually able to like consistently do the thing you want, you actually have to chop down the prompts more. Like you're asking for the LLM to do too much. Like, The LLMs are capable of maybe like 120 IQ level work right now. And if you're, you know, giving it like too much in the context window and asking it to output too much, it's just like, uh, do it in steps, right? This is literally what O-1 and, uh, you know, reasoning already does using chain of thoughts. The interesting thing and weird thing that I'm tracking right now that I'm a little bit worried about is it is entirely possible that the next generation of, uh, large language models from Anthropic and OpenAI and Meta. Are, you know, one and then two orders of magnitude, uh, in num— in like compute and number of parameters. And then we just get like, you know, the Dario Amodei scaling laws paper specifically says like, you know, these big large language model, uh, AI labs are going to spend a billion and then $10 billion on the next generations. And like, if we get like continuous step function improvement, like the trickiest thing is, like all the stuff that even what we're doing as founders are doing, like, you know, maybe the models will just do for themselves. And, uh, you know, I think that my hope is that, um, you know, there's not enough agency in these things. And then, you know, maybe what will happen, what, what we hope and what we think will happen is that the people who establish brands and moats and evals right now when the next generation models happen, their cost structure just comes down by like 10x. And then they already have like the revenue and the market captured in some way that they're like sort of beating the software incumbents that they're going up against. They themselves have like tens to hundreds of millions of dollars in revenue. And then like basically the standard moats apply, like look at Porter's Five Forces and you're like, you know, uh, your data, you have way, way better access to data. Your, your systems are way smarter. It costs a lot of money to take a risk and switch off of those systems. And then you underst— you have a sales force that is way better and smarter. And so if you take all, you know, that's sort of the gambit right now. Like if you do prompts and evals, you are pulling forward the future by a few years. You're trying to like grab as much land as you possibly can right now. And, uh, and then from there, like, hopefully you can hold on and build the next Microsoft or the next Salesforce or, you know, the next $10 to $100 billion company.

Was it obvious when Sam Altman left, uh, YC to do this?

I wish I was smart enough to predict that this was what was going to happen.

Did you think that he was going to be the man like he is now? Like, you know, Paul, Paul Graham has that essay where he's like, here's 5 people I'd bet on and one of them The fifth one was like a 25-year-old Sam Altman, and the other three were like, you know, uh, Bill Gates and Steve Jobs. And then he said this kid. So did you, were you like, whatever Sam does is gonna be a generational company? Or were you like, eh, it might be, it could be mildly successful.

I mean, I think it could be. I bet in full transparency, like I'm gonna be super honest. I was like, man, I don't, I don't quite get it. You know, I, he, uh, when he was sitting in my, my shoes as president of YC, uh, he'd come back and I remember sitting in the group partner room and he'd say like, well, this is what Elon was talking about and this is what Larry was talking about. And you know, they were all talking about AI and they were, you know, sort of talking about these sci-fi scenarios. Um, and I'm like, man, I saw Terminator 2 too, but that was a movie. Okay. And, uh, I think that I'm changing my tune at this point. Like it, it, You know, I wish that it was smart enough to be, you know, actually totally a believer. You know, I'm a believer now. Um, do I believe that we're going to end up in like a doomsday scenario?

Like, I still don't really believe that there's that much agency in these things, which is sort of your judgment now is questionable, you know, first Peter, first Palantir, now OpenAI. You know, thank, thank God that you've like already made so many other successful investments that I know that Gary's got good judgment and like can predict the future a little bit because, uh, I'll come around eventually.

You know, like I'm just a year off.

Yeah. Yeah. He was Palantir employee 10 after, after he said no to the co-founder. Well, he ended up there. Right.

So, I mean, now we're, you know, we are full believers in large language models, but I, you know, I think the thing that's a little bit lost to the sands of time is that, you know, what, what Sam was absolutely right about was, um, jumping on this thing that the smartest people in the world were already talking about like 8, 9 years ago. And then he built it, you know, he invested the money. He, um, you know, brought together the people. It was no small task to bring together the smartest AI researchers in the world. And then not only that, it was like semi-miracle. You know, you could ask, why didn't Meta do this before? Why didn't Google do this before? Why didn't Microsoft Research do this before? Like, all of those people had virtually limitless access to resources. And, um, to weave this back together, I think there is something really terrible and limiting about, like, giant organizations that are not capable of evolving and making smart decisions. I mean, that's what big tech is like. They're sort of like big daycares for the most technical smart people in the world. And it's like, here's— here, do you know, play some volleyball, we'll cook you lunch. Like, here, you know, here's— do your laundry at work, you know, like You know, let's totally infantilize the smartest people in the world. And like, that's a waste of time. Like, what needed to happen was, um, the ability, you know, uh, one of the things that got explained to me why, uh, it was OpenAI that came up and like invested, you know, millions first and then tens of millions and hundreds of millions into the large language model and, you know, transformers for language. Um, if you were at Google, you needed to work this bureaucracy. It was like highly political, you know. You had hundreds of other researchers, all of whom were all as smart as you, and you had a finite sort of limited set of compute. And, you know, what you would find is that, um, Google would do really breathtakingly amazing work, but then you'd have like 30 different authors You know, and if you'd read the paper, there'd be like these sort of like strange things that seem like it was a little bit tacked on.

Yeah.

But that's the only way you could get the compute resources to actually pull off a training run on like the model that you wanted to do. Like you had to do some weird things to accommodate like people who were gonna commit their training resources for their like random little thing. And so it's funny to see that. You know, human progress is more or less a little bit, uh, impeded by the bureaucracy and lack of governance and lack of agency that large organizations just sort of like impose on people. Like, you don't have a choice.

Isn't it crazy that it was easier for them to invent and discover Transformers and this large language model breakthrough than it was for them to navigate the bureaucracy to actually like create a product out of this or do anything useful with it? Like, that should tell you kind of everything you need to know.

I mean, the wild thing is like, even with an OpenAI, like they did everything right. Like GPT-3 was very impressive. 3.5 was even more. 4 was incredible. But like, they were going to just put it behind an API and like explicitly for many years, OpenAI was supposed to be a, you know, we're a research lab, not a product company. And, uh, you know, it took the relatively Herculean efforts by product people inside OpenAI to even release, um, you know, ChatGPT, which turned out to be the greatest consumer launch of maybe the, in the history of consumer launches, actually.

Hey, you said, um, and I, and I know we gotta wrap it a second, but you said, um, you keep talking about this thing. It's pretty amazing about, uh, talking to the smart people and like, what, what are they, what are they talking about? Uh, you have access to more of those types of people than probably most anyone on earth, what's the conversation right now? Or is it, is it all AI still?

Yeah, I mean, ultimately that's sort of the number one thing that we're pretty excited about. I mean, that, you know, a few people who are really smart can sit in a room and, I mean, make a thing that literally does what humans do all day. You know, I, I would make a pitch that often the best things that could be automated are actually very rote. They're sort of the knowledge worker equivalent of like hand looming a carpet or something. It's like all I do all day is look at an email box, see if someone paid their bill, reconcile that against a ledger in QuickBooks. Like, can you imagine, you know, millions of people, like tens of millions of people on the planet, like that's their 9 to 5, you know, like that, that is not a good use of a very smart human being, uh, generally speaking. And like, those are sort of the things that are most easily replaced by really great prompting, great evals. And frankly, I think, um, you know, there are going to be 20-person software engineering-focused startups that basically turn, um, you know, billions of dollars in payroll into billions of dollars in software revenue. And Hopefully those people go on to do much smarter and much more interesting jobs than like passing butter all day.

That's amazing, Gary. Uh, if you have one minute, can you tell the spoon bending story? I think this is the, uh, a very powerful story that you have, uh, the spoon bending party. I don't know if you know this.

Oh God, it's hilarious. I mean, uh, I, you know, I've been to Burning Man a few times. I love Burning Man. It's my favorite place to take a break from consensus reality and realize that it's all made up, but you get to make it up. And, you know, there's a lot of like weird fringe stuff, a lot of like '60s, like esoterica, you know, you're just wandering around the desert on a bike with your buddies. And, you know, I think we randomly walked into a spoon bending party where they were, you know, it was a huge group of, I think it was almost like a magician at the front, like sort of helping people. You know, the shtick was that they were gonna teach you how to bend a spoon with your mind. I was like, okay, that's cool. Like, let's go check it out. They passed out these spoons, um, and then it was like a magic trick. They said, uh, okay, now bend your spoon with— like, look at your spoon and bend it with your mind. And then in between, they would say like, okay, well, sometimes you need to use your hands and like warm up the spoon to bend it. And then basically, uh, I think What you're supposed to do literally is just bend the spoon with your hands. And that's what, you know, and then what's funny is by the end of it, um, half of the people, like I had, I had a bent spoon cuz I was like, oh yeah, like they said like, oh, I should, you know, sort of warm up the spoon with my hands and bend it with my hands to get it started.

Just to get it started.

Yeah. Just, yeah, just to get it started. Right. And then by the end, like half of the room had bent spoons and the other half didn't. And the ones that didn't were like, how did you do that? That's absolutely incredible. And I was like, guys, it's a bit. It's a bit like you have to, you know, first know that you can bend the spoon with your mind, but then you bend it with your hands, guys. Like, that's literally like the shtick, actually. And it's a little bit of an allegory for, uh, you know, bias to action and what you're supposed to do in real life. Like, real life is sort of like this exercise. Like, they say, like, you can go change the world and you can do it with your— I, you know, with your very ideas. But like, You know, that alone is not enough. Like you actually have to go with your hands and go talk to users and, you know, sit and build a thing. And so, you know, don't forget the second half. And the wildest thing is like more than half of the room could not do it. And so if you are the person who's able to bend this, you know, bend the spoon with your mind, like, guess what? You also used your hands. Like you gotta do that. And you know, and actually like that's. Maybe like the greatest gift.

It's just like, dude, Gary, you're the shit, man. You are the best. I love talking to you. You make me feel really great. You give me energy and we're very, very, very thankful that you took the time to do this.

Thank you guys so much for having me. Always great to see you guys and, you know, congrats on all the success and excited to see where you and your community go. I mean, you know, we're, we're all building the future out here.

Well, thank you for saying that.

Take care, Gary.