[0:00:00] When we started out, we didn't seem like we were going to be successful at all. OpenAI had a billion dollars and all of the star power, and we had seven co-founders in COVID trying to build something, and we didn't know if we were necessarily going to make a product or what the products would look like. One thing that's interesting to look at is just that
[0:00:21] Humanity is on track for like the largest infrastructure build out of all time. Tell us about the early days of Anthropic. So you had a general idea of this sort of like long term mission that you wanted to do to, you know, not destroy humanity. But like, what did you actually work on for the first year? How did that converge on an actual product? Welcome back to another episode of The Light Cone. Today we've got a real treat, co-founder of Anthropic, Tom Brown. Excited to be here.
[0:00:53] Tom, one of the things that a lot of the people watching would love to figure out is you got started in tech at the age of 21, fresh from MIT. How does someone go from that in 2009 to...
[0:01:08] Literally co-founding something as important as Anthropic. Summer 2009, linked language. Two of my friends had started that out. I think they had seen one of our other friends, Kyle Boat, kind of do a YC company. And so it was in the water that that's a thing that we could try to do.
[0:01:24] They started out, I was the first employee. Back then, yeah, you guys let me join for all the dinners and stuff like that too. I could have instead gone to like a big tech company or something like that. And I think probably just as a software engineer, I might have learned more software engineering skills. But I think by being there with...
[0:01:43] The other co-founders without anyone telling us what to do. Basically, we had to figure out how to live. The company would die by default. I think in school, there was a lot of a feeling of more of people would give me tasks and I would do the tasks. It's kind of like a dog waiting for food to be fed to them in their bowl or something like that. And I think for that company, it was more like wolves and we have to hunt our food. Otherwise, our kids are going to starve or something like that. I think that that mindset, I think, has been the most valuable mindset.
[0:02:13] shift that I've had for trying to do like bigger, more exciting things. Yeah, big tech just teaches you to work at a big tech company, whereas it's much more fun to be a wolf. Yeah. How did you go from like, so working at friend startup to then you started your own one? So linked was, we ran the company for a bit, I ended up going back to school afterwards.
[0:02:37] And then when I left school, I went to this company, Mopop. The mobile advertising thing, right? Yeah, yeah. I was like the first engineer there. I was like, okay, I want to be a wolf, but I was really bad at programming also. I was very, very struggling as a software engineer. I know I want to do more, but I don't know how to do it yet. And so I think that was kind of like an experience getting to scale something. Winter 2012, one of my friends who was my smartest friend from...
[0:03:04] College pitched me on let's go and start a YC company. We did at the time solid stage. This was before Docker existed. And so the idea was try to make it easier to do DevOps, but Docker doesn't exist. So it's going to be a more flexible Heroku, which basically meant a more complicated like Heroku. And so I remember we interviewed with you guys. I think folks didn't really understand what we were trying to build. I think we didn't really understand what we were trying to build that much. When you're trying to do something new, that's actually sometimes...
[0:03:33] Yeah, I think we were an outlier there because we like did our interviews and then we got called back driving back to San Francisco and TLB had written on the board like an angry frowny face. What are you actually going to build? And so he like wanted us to explain that. I guess we explained it enough or he was just like, these guys still don't know what they're doing, but maybe they'll figure it out. Halfway through, I kind of felt I still didn't actually understand what we were going to build and how.
[0:04:00] We would attach a mission to it that I wanted to work on for my whole life. And so I left. PG actually introed me to Michael Waxman, who was the Grouper founder. So Grouper was a dating app, only it was novel in that you had, what, three guys and three girls. This was before AI in a lot of ways. So there was a team of people who would manually link people up, right? And they'd meet up at a bar and shenanigans would ensue.
[0:04:30] Reliably shenanigans. People didn't always have a great time. I think you went on a couple of groups. The pitch for Grouper for me, for why I was excited for it, was just I was an incredibly awkward kid. What I wanted to do was to basically have a thing that lets awkward people like me go out and...
[0:04:51] Talk to other people. For me, talk to girls and feel like I was safe doing it with my friends around and stuff like that. And so I think who are going to be our employees was important. I did all of our engineering interviews. We would take someone. The only person who went on more was Greg Brockman. I think he had a phase where every single week.
[0:05:11] He would go and post on Slack or HipChat at the time or whatever. Because he moved to New York and he was hanging out at the Recurse Center during this period, I think. Oh, I think he was at Stripe. Maybe for part of it he was at Recurse, yeah. But he also had, I think, just a phase where he would just, at Stripe, he would just post in their thing every, like, I'm going on a group or who's going for a whole year.
[0:05:31] So I ended up being close with Greg, which ended up being a connection to the OpenAI. What was the journey like? Because you started as, you just graduated from MIT CS. You were 21. You became first an early employee for all these YC startups. Then you started your company just a couple of years later. And what was the path for you to eventually become the co-founder of Anthropic?
[0:05:55] Long path, but it's pretty impressive. How did you get there? I mean, it sounds like getting in touch with Greg at that moment was a big moment. And then you were one of the first couple dozen people to join OpenAI as a result. Yeah. So I left Gruber 2014, June 2014, and I joined OpenAI, I think, a year later. I tried to build up courage to make the switch to be a...
[0:06:24] Try to learn AI research. At the time I was like, okay, it seems like sometime in our lifetimes, we might end up making transformative AI. If we do, that would be the biggest thing. Maybe there's some way that I could help out. But also I got like a B minus in linear algebra in college. And so it seemed like at the time you needed to be just top superstar in order to try to help out with that at all. And so I think I had like a lot of uncertainty about whether I would be able to.
[0:06:49] Help. And also I'd had some success with startups. And so a lot of me was just like, rather than trying to retool at this, like I could try to do another startup or something like that. I feel like in that period, going to work on AI research, which is not seen as like a serious, like not like a practically serious thing to do. And you're in a world where it's like people try and build companies and do these like really practical things. So what did your, were your friends like, oh, that's really cool. You're going to work on AI stuff or was it?
[0:07:12] Not really. I think my friends were like, that sounds weird and bad. It doesn't seem like AI safety is a thing we should be aware of. Overpopulation on Mars doesn't make any sense. And my friends were also just like, I don't know if you're going to be good at that, Tom. I think that for that reason, I think I didn't try very hard. I kind of flip-flopped on it for six months trying to build up courage to do it. And what were you specifically at this point? You're reading research papers. What does it look like?
[0:07:41] Yeah, so first I was just kind of hanging out. I built like an art car for Titanic's End and stuff like that. Oh, that was fun, yeah. So I spent like a whole summer, like three months after Grouper doing that. Because honestly, I was like kind of burned out for Grouper where...
[0:07:55] I don't know, startups, like the highs are high, like the lows are low. And we weren't working at the end. Our business wasn't succeeding. Our revenue was going down. But my main job still was like recruiting engineers. And so I had to like pitch them on this dream that I'd had, but I like no longer really believed in. It sounds like a death march. Yeah. And so I was super burnt out. And I was like, okay, Tom, like chill out, do some yoga, like do some CrossFit, like build an art car. What was the hindsight, like, you know, hindsight 2020, what's the retrospective on like Grouper obviously attracted all these really, really smart people.
[0:08:25] The graphs were up and to the right, and then it flatlined and maybe started declining. What happened? I think that when we started, the competition was like OKCupid. It was all web-based. All web-based. The main problem that I think we were solving was it's hard to go and put yourself out there and go talk to someone new, and they might just be like, I don't want to talk to you. You seem weird. And so we solved that by just blind matching. Tinder came out while we were.
[0:08:52] And then, yeah, like, when did you get serious about AI and just how did you approach it?
[0:09:16] Three months of like kind of playing and having fun. And then I ran out of money also. I had like my personal runway. I ran out. And so I was like, OK, I think that I'm going to need six months of stealth study to have a shot at getting a job. At that point, it was DeepMind or Google Brain were the two places to do work there. Or Miri. Miri was the third one that I was looking at. So I was like, if I want to help out.
[0:09:40] With that, those are the three places to look at. I don't have any of the skills yet. I need six months of self-study to feel like I would not be a drag on them and actually be helping instead. Can you maybe explain a bit what was a self-study like? Because I'm sure there's a lot of software engineers right now in their 20s that are looking to become AI researchers.
[0:09:59] What was that six months like? Even though, as you said, you had gotten a B-minus in linear algebra, which is like core. Might have been a C+. I should check. I'm going to keep telling you. But that's pretty impressive where you got to. Yeah, yeah, it turned out okay. First, I did a contract actually with Twitch and earned enough to cap that six months of runway. So I did a three-month contract with Twitch, and then I made a plan to self-study. I don't think it's the right plan now for people, too, at least in 2015.
[0:10:27] What did it look like? It was like take a Coursera course on machine learning, try to solve some Kaggle projects, read linear algebra done right. And I had a statistics textbook. I think I had YC alumni credits. And so I bought like a GPU and I would like SSH into the GPU to like work through my courses for it. And this is right after, yeah, it was already after I like snack, right?
[0:10:54] This is after AlexNet, yeah. So I was mostly doing image classification stuff that I was trying to learn was like the thing that all the courses would teach you to do. How did you get the OpenAI job? Because you were one of the few engineers who was mostly researchers and they had a pretty stacked team of researchers. I messaged Greg as soon as OpenAI was announced and I was like, I'd love to help out in some way.
[0:11:16] I got a B minus in my linear algebra. But I know some engineering. I've done a bit of distributed systems work. If you guys need help, I'm happy to mop floors if you guys need. I want to help out, however. And I think Greg was like, yeah, I think there's a paucity of people who... He said paucity, too. I was like, fancy word there. There's a paucity of people who know both machine learning and distributed systems. So yes, you should do that. I think he introduced me to Peter Abiel also to help me put together a little course for myself, too. And then I...
[0:11:45] Checked in on with him, I think, every month or something. And then after a couple months, he was like, oh, we actually have a project, which is we need to put together, we want to play games. Can you help make StarCraft environment? And so I joined to help them with the StarCraft environment. So that ended up, I think, getting my foot in the door. I didn't do any machine learning work with them for the first nine months that I was there, basically. And what did OpenAI feel like at...
[0:12:13] At this point, had it raised much funding? Did it have an office? Did it feel like a startup? So it was in the chocolate on top of the Dandelion Chocolate Factory. This is after Greg's apartment. After Greg's apartment, yeah. So right after Greg's apartment in the Chocolate Factory. When it kicked off, it was like a billion dollars of committed funding from Elon. It felt like it was very solid. The other interesting milestone for you was...
[0:12:38] Yeah. So I ended up...
[0:12:57] Working at OpenAI for a year, left, went to Google Brain for a year, came back, and then GPT-3 was 2018 through 2019 was like building up to GPT-3, which exactly, as you said, was like scaling things up. I think that like Dario had seen the big trend of scaling laws, basically. You published a paper for that. Yeah, yeah. And that's like a pretty important paper that now has withstood the test of time, and we're living now the dream of it.
[0:13:22] Definitely seeing that line of reliably you get more intelligence if you spend more, compute with the right recipe was the main thing that was, at least for me, was like, this is a thing that's happening now. Because you could look even at the time, we weren't spending very much money on the training jobs at the time. And you could see that there was scaling there. And then also, Danny Hernandez did a paper at the time that showed how much cheaper algorithmic efficiency was making stuff over time, too. And those two things stack.
[0:13:51] Together, that was like, oh, wow, we're going to get a lot more intelligence over the next few years. So it was noteworthy and surprising when you saw it. Yeah. And I think the thing that seemed the weirdest to me is like, I'm not a physicist, but like all these physicists were doing this stuff. The like original scaling laws paper, just like very straight line over like.
[0:14:11] 12 orders of magnitude. I'm just like, 12 orders of magnitude is like, just like a stupidly large amount of, I've like never seen anything go over 12 orders of magnitude. That convinced me to definitely pivot all of my work into scaling, which I hadn't been doing before. Can I ask a like kind of layperson question? I mean, is it fair to say that the scaling law?
[0:14:31] Yeah, so I think in physics, scaling laws hold all over the place, which I didn't know at the time.
[0:14:48] Within physics, there's a whole field called phenomenology that basically looks at various aspects of the world and then does those types of fits. And they find these power law distributions all over the place. This was, I think, the first one that I had ever seen in a computer science adjacent thing, which I think was interesting.
[0:15:10] Surprising. And at the time, people were mad about it. They actually were like, you're throwing money at GPUs or you're just wasting money. This is very wasteful. People are still mad about that. Yes, different people now, but still people mad about it. Yeah, I guess the researchers were mad at that too, where it's not elegant. You're just brute forcing it. The gesture cap, stack more layers.
[0:15:35] Which I think Anthropic's slogan, I think, is do the stupid thing that works. That was a thing where this was very clearly the very stupid thing that works. Can you tell us then how you ended up collecting the last Infinity Stone? With Anthropic? Yeah, with Anthropic. There's very few people in the world that have basically worked at OpenAI, DeepMind, and Anthropic, and you were part of the team that spun off from GPT-3 and then started Anthropic.
[0:16:02] How was that jump? There were two teams there. There was the safety org and the scaling org were the two orgs that reported into Dario and Daniela. I think we had just like worked together extremely well. One thing I think that was great, both at OpenAI and at Anthropic was just like we had a culture where like everything is on Slack, 100% of things on Slack. And within that, all public channels, great communication.
[0:16:28] I think that that group also was the group that took the scaling laws the most seriously. Where it was like, okay, this actually is going to be transformative. There's going to be a handoff where humanity will hand off control to transformative AI at some point. And hopefully they'll be aligned with us and that'll be a good transition that goes well. But it might not be. The stakes are incredibly high. And so I think that group was very focused on how do we make sure that that's taken seriously.
[0:16:56] And that we've built an institution that can handle the weight of that. That ended up being the core group that left to join Anthropic. And I think it wasn't clear at all to me that that was the right thing for the world at the time. In hindsight now, it seems like that was a good choice. I think what was kind of cool then, too, is when we started out, we didn't seem like we were going to be...
[0:17:17] Successful at all. OpenAI had a billion dollars and all of the star power. And we had seven co-founders in COVID trying to build something. And we didn't know if we were necessarily going to make a product or what the products would look like. And so I think that what was interesting from that, too, is that all of the initial people who joined were there for the mission, too. They all could have worked somewhere else for more prestige, more money. People would have known what they were doing, et cetera. What's data at OpenAI?
[0:17:46] Exactly. Yeah, exactly. That's been an interesting thing then that I think has been like the key to like letting our culture or like let our org scale. We're like 2000 people now, but we still have a thing where it doesn't seem like politics have creeped in. And I think a lot of that is like the first hundred people all were just there for the mission. So like if something starts to go wrong, they'll like raise their hand and be like, it seems like this person might not be acting for the mission.
[0:18:09] YC's next batch is now taking applications. Got a startup in you? Apply at YCombinator.com slash apply. It's never too early and filling out the app will level up your idea. Okay, back to the video. Hey, tell us about the early days of Anthropic. So the time you broke off from OpenAI, you had a general idea of this sort of like...
[0:18:32] Long-term mission that you wanted to do to not destroy humanity, but what did you actually work on for the first year? How did that converge on an actual product? So first year...
[0:18:44] The main thing that I tried to do was just build the training infrastructure that we needed to train a model and then get the compute that we needed to train the model. Those were like my two main projects. All the other things that you need to do when you're like starting up a company too. So like set up a Brex account and like, I don't know, like all of that stuff. We started out with seven co-founders. Within like a few months, I think like 25 folks from OpenAI.
[0:19:09] Overall had joined. So we had like a pretty substantial team that like already knew how to work together too. And so that helped us get up and running faster. And at what point did you launch the first product? And when did things begin to actually start working? So the first product that we launched was after ChatGPT. We had like maybe nine months before ChatGPT, we had a Slack bot version of like Claude 1. Oh yeah, we had that in the YC Slack actually.
[0:19:37] Yeah, I remember Tom Blomfield adding all of you guys to it also. It was really cool. And then I think that at the time, though, we didn't know whether or not we wanted to launch it as a product. We didn't know if doing so would be good for the world at the time. I think we hadn't really thought through our theory of impact that much for how we actually will make stuff work well. Plus, I think in hindsight, if we had tried to launch it, we wouldn't have had the serving infrastructure.
[0:20:03] And I think because we weren't sure whether or not we wanted to, we hesitated for too long on building that infrastructure, which I think is learning for me. I mean, at this time, ChatGPT had not launched yet. ChatGPT hadn't launched. And so I guess we didn't know that it would be a big deal, too. This is around the pandemic, 2022. This is summer 2022, yeah. And then ChatGPT launched fall 2022. And then we relaunched our API after that, and then Claude AI after that also.
[0:20:33] I think it didn't seem like it was working basically until Claude 3.5 and coding. I think really through that whole time then until about a year ago, it seemed like it wasn't clear that we were going to end up being a successful company. We actually saw that in the startups because we kind of get a bit of a vibe check in terms of what is the preferred model for startups. So all of 2023, OpenAI, OpenAI was the response. Then things started to turn in 2024.
[0:21:02] is when we saw Claude 3.5 and especially Sonnet was starting to get market share per se in the YC batches going from single digit to at some point like 20 and to 30%. And especially for coding became the default choice, which was very interesting. Can you tell us about how that emergent behavior and the spikiness on that particular skill? Must be 80% now or 90. Yeah, for coding even more. Especially now Claude code.
[0:21:29] What was that? Was that on purpose or just kind of happened? I think that we invested more in trying to make the model really good at code because we wanted the model to be good at code was one thing. Then you did it. And then I think seeing the reaction of everyone to it was like, okay, yeah, let's go much harder on that also. And this is before 3.5 signed it. You'd already invested enough in coding to realize that that was really promising. And you said, I decided to double down.
[0:21:57] I think this really was like individuals within the org being like, we want to do coding before 3.5 Sonnet. And then when we saw 3.5 Sonnet's really good product market fit, that was a good signal to go for that. And did you guys know, the day that you guys launched 3.5 Sonnet, did you know that you had something really special and this was going to be the turning point for the company? Or were you as surprised as OpenAI when they launched ChatGPT and it just unexpectedly took off? Yeah, I wish that we had more foresight on that. But no, I think it was surprising for us too.
[0:22:27] Like how big of a deal it was. And then I think 3.7 Sonnet also like surprised us by how much it unlocked like agentic coding. I think for each of these things, yeah, we move quite fast in rolling them out. And so we really often don't know what the results are going to be there. I think it's what made a lot of these coding agent startups work. I mean, there's a crazy story of Replit winning, going to a hundred million in just 10 months, right? There's Cursor, of course, a story and all built
[0:22:56] I think that all of those things have been surprising to me. And then also just like in my working with Claude too, like I think I continue to be surprised by like the type of stuff that it can do. And I do think with each one, there's like more stuff that kind of unlocks. One of my friends was telling me that she had some...
[0:23:13] Oh my god.
[0:23:43] We keep getting surprised by stuff that Model has memorized, all the hex tables it can think through, try to work through it. I think we're going to continue to be surprised by that sort of stuff, too. If you poll the YC founders, they prefer using Anthropic models for coding by a huge margin. It's much larger than what you would predict if you just looked at the benchmark results.
[0:24:02] So there seems to be some X factor that makes people really like these models for coding. Do you know what it is? And is it intentional in some way? Or it just came out of the black box somehow? I think that the benchmarks are easy to game, where I think that all the other big labs, I think, have teams where their whole job with the team is to make the benchmarks scores good. And we don't have such a team. And so I think that that is probably the biggest.
[0:24:29] You don't teach to the test. We don't teach to the test. Because I do feel like if you start doing that, then it has weird, bad incentives. Maybe we could put that team under marketing or something like that and then ignore all the benchmarks. But I think that that's one reason why there's some train test mismatch there. So the evaluations are more qualitative internally? We have internal benchmarks, yeah, but we don't publish them.
[0:24:52] And is it the internal benchmarks that the teams are really focused on improving? That's right. Yeah. So we have internal benchmarks that the team focuses on improving. And then we also have a bunch of tasks. Like I think that accelerating our own engineers is like a top, top priority for us too. And so we do a ton of like dogfooding there to make sure that it's helping with our folks too. Going back to GoldenGate, Claude, there's a lot of sort of the interpretability seems like it's a big part of it. And then most people would say that.
[0:25:20] You know, Claude's personality just feels better. And then how do you sort of at once be very quantitative, but then also, you know, build evals around personality? The evals for personality are kind of complicated, too, for like, how do you tell if like Claude has like a good heart or something like that? It's like hard to know. But I do think that that's like Amanda Askell's team's mandate. I think she describes it as like being like a good.
[0:25:45] World Traveler, where Claude goes and talks with all sorts of people from different backgrounds, and each of the people should come to that being like, I feel good about this conversation that I've had. Interprely, I think, is a long-term bet, right? Where it's like, right now the models aren't that scary, but at some point they're going to be more scary. And so I think the hope there is to have some ability to know what's actually going on under the hood when it becomes more intense. Then more recently, Claude Code's been a real success. Can you talk us through like...
[0:26:13] How did that project get started internally? And again, was it like a, did you like know this time it was going to work or was it a surprise?
[0:26:20] Claude Code was an internal tool also. So like try to help out our engineers within Anthropic that, yeah, Boris had like hacked together. There's an internal Anthropic engineer wanting to build it for themselves. For other internal engineers, yeah, for him and other internal engineers. And then I think, yeah, I think we definitely didn't know that it would be successful out there. And I think to some degree, like we really had fully just bet on the API before that with the intention being like,
[0:26:50] There's so many startups out there with so many good ideas. Who are we to figure out what the right product is to build on top of this stuff? Everyone out there is going to build better stuff than us. And so put all of our effort into just making the best possible API. And I think that this surprised me as like, okay, we actually were able to make something that as a product was better than the other products out on the market for this agentic use. I have some theory that like...
[0:27:17] Part of that came from a mind shift of seeing Claude as the user for this thing, too. For LinkedIn, we were trying to build things. For teachers, we're our users. For Grouper, it was single people in New York, mostly, I guess. For this, I think really the users are the developers, but also I think the users is Claude. It's like, give Claude the right tools that Claude can actually...
[0:27:40] Do that effectively. Help Claude get the right context to work effectively. This team was like the most focused on Claude as like a user, which I think makes sense. You guys would understand Claude the best. Yeah. I do think that that's a place where like startup founders though, like can, can do that too. And I think that that's, that's probably a rich vein for people to like make tools that are better for models as users.
[0:28:04] That's the perfect anthropomorphization of the LLM itself. The agent is one of the stakeholders, one of the users that you would go after and try to empower. Yeah, totally. Which actually makes a lot of sense why you guys actually got MCP to work, to do tool calling, because a bunch of other labs had tried to do something, and the standard that really took off was yours.
[0:28:29] Yeah, I think that seems like a similar one, too, where it's model-focused. Going back to Claude Coates, success is really exciting. It's also scary for Cursor and other companies that have built on top of the API. What's your advice to founders building products? How should they think about it?
[0:28:45] Building on the API, but also worrying about Anthropic or one of the labs building something better than they can build. I think I was kind of surprised that ClaudeCode, we did build a thing that was the best in the market there too. It's not super clear to me.
[0:28:59] What the big advantage was for us for Claude code besides more empathy for Claude. I think that's actually a really interesting insight. Like it seems like the thing that, yeah, you were building for a specific user that you knew really well that other people wouldn't have thought to build for versus like you had some like intrinsic technology advantage. Yeah. Like I think a startup could have done that same thing too, right? Yeah.
[0:29:20] I think we're the most developer focused lab. I think we're the most API focused lab too. So I think we want to make sure that we have the best platform for people to build stuff on because this thing is growing so incredibly quickly. We're not going to be the fastest at figuring out all the ways that we need to empower Claude to do the work that connects Claude to the entire human business. That's like human world is all designed for humans, but we need to get the models to be able to be productive members of.
[0:29:48] Are there ideas or areas you would love to see developers building in? Or areas you think are underappreciated right now? Yeah. Claude code is like how do you get Claude to be a useful pair programmer, kind of. Or like junior engineer. You've got like a SWE level...
[0:30:09] Two or three or something like that that you can work with or like very spiky because also it can do the like weird disassembly stuff that like a super high level suite would struggle with. Less good at knowing what type of work to do. Needs kind of a lot of handholding. Needs a lot of context from it. That's like one very particular subset of work that can be done. If you look at like all the stuff that happens in businesses besides that, it's like a very tiny fraction of like all the work that's done in.
[0:30:38] businesses that a smart person who knows how to code and use lots of tools but doesn't have that much context yet would want to do. So I think finding ways to coach Claude or coach whatever model to do useful tasks for businesses seems like there's just a huge, huge space there. So Tom, a big part of your job is like...
[0:31:03] Owning all the compute infrastructure that makes Anthropic work. Can you talk about what is the compute infrastructure behind this giant thing now? One thing that's interesting to look at is just that humanity is on track for the largest infrastructure build-out of all time now. Is it going to be larger than the Apollo project? Larger than the Manhattan project? It'll be bigger than both of them next year if it keeps on the current trajectory, which is roughly 3x per year increase in spending on AGI compute.
[0:31:32] Which is just bonkers. Yeah, like 3x per year is wild. I think it's going to keep up on the 3x per year trajectory. It's already locked in for that for next year. And then it's a little bit open for 2027. I mean, anecdotally, internal to YC, we can't get enough, you know.
[0:31:51] Credits across all of the top frontier models, including Claude. So you got to help us out a little bit. I mean, everyone's bottlenecked. It's like, give me more intelligence. I can't have enough. And I know you guys have been looking at more hardware startups also for more accelerators. I think that we will see more accelerators coming online to 2027. That's a good space. Also, data center tech, I think, is a big one. Where are the bottlenecks for you guys now?
[0:32:18] Getting enough electricity, getting enough GPUs, getting construction permits. Power. People are using jet engines to get power. That's nuts. Overall, for the build-out, I think power is going to be the biggest bottleneck, especially power in the U.S. We want to build in the U.S. That's one of our biggest policy goals is to get the U.S. to build more data centers, permit more data centers, make it easier to build. Is the answer renewables or is it nuclear?
[0:32:45] I definitely, I feel like, yes, yes, all of those things. I wish the nuclear was easier to build. Anthropic is the only major lab that uses not just one kind of GPU, but the GPUs from three different manufacturers. Can you talk about that and how that strategy is played out? Yeah, yeah. So we use GPUs, TPUs, and Tranium.
[0:33:06] Downside of doing that is that we split our performance engineering teams across all of those platforms, which is a ton of extra work. The positive thing is it gives us the flexibility to both one, like soak up that extra capacity because they're there.
[0:33:20] It just is more of those altogether than just one. And then two is we can use the right chips for the right jobs, where some chips will be better for inference, some chips will be better for training, and we can match the right chips to the right jobs. So yeah, I think that's kind of the trade-off there. I guess one cool thing is just connecting the dots through your career and how all of this compounded, because you were the one engineer building that change of the architecture from TPUs to GPUs back at...
[0:33:48] OpenAI that got GPT-3 to actually scale. And now you're in charge of that at a much, much bigger scale years later. I don't know if that kind of connected that for you. The big move from TPUs to GPUs at OpenAI, I think, was partly driven just that PyTorch was a better software stack on top of them than TensorFlow on top of...
[0:34:09] And I think that that then unlocked fast iteration where like if you have like a good reliable software stack, then you can experiment quickly, just like build a whole system that works. I think that that's the thing that we really strive for now at Anthropic too is a challenge of having many more platforms is that it's harder to write all the good software. I think building the muscle.
[0:34:30] Knowing how to build that software well so that all of the people who build on top of that low level can have a great experience with it is the most important thing there. Do you have advice for kind of like a younger Tom version of yourself who now you've seen and went through this crazy journey? If someone was you back in the 20s living today and they wanted to arrive and join the AI revolution, what would you say to them? Very specifically, something we see from a lot of, hear from a lot of college students at the moment is they...
[0:34:59] They don't know if they should stay in college. Are there going to be jobs for them? How is the world going to change and what should they do? Taking more risks, I think, is wise. And then also trying to work on stuff where your friends would be really excited and impressed if you did it. Or a more idealized version of yourself would be really proud of yourself if you succeeded at it. I think it's probably the thing that I would try to tell a younger version of myself. More intrinsic, less extrinsic.
[0:35:29] Don't chase these other credentials and getting the degree or working at Fang. Those are just irrelevant as of today. Yeah, exactly. That's all we have time for today. We'll see you guys next time.