LIVE: Sam Altman of OpenAI on Building the ‘Core AI Subscription’ for Your Life

[00:00] Hi, and welcome to Training Data. We are mixing it up for this week's episode and dropping a conversation that was filmed live at Sequoia's annual AI conference in San Francisco with OpenAI founder and CEO Sam Altman. Sam is interviewed by our partner Alfred Lin. [00:14] We hope you enjoy this special conversation with Sam about the Genesis story of ChatGPT, his predictions for agents in scientific discovery and robotics, and more. And stay tuned for a few more special AI Ascent drops on our podcast feed later this week. [00:30] Our next guest needs no introduction, so I'm not going to bother introducing him, Sam Altman. I will just say Sam is now three for three and joining us to share his thoughts at the three AI sense that we've had, which we really appreciate. So I just want to say thank you for being here. [00:48] That's right. Oh, that's right. Say that again. Yeah, this was our first office, so it's nice to be back. [00:55] Let's go back to the first office here. It started in 2016. Yeah. 2016. We just had Jensen here who said that he delivered the first DGX1 system. [01:07] Over here? He did, yeah. It's amazing how small that thing looks now. [01:10] Oh, versus what? Well, the current boxes are still huge, but yeah. [01:14] It was a fun throwback. [01:15] How heavy was it? [01:17] That was still one you could kind of like lift one yourself. [01:21] He said it was about 70 pounds. I mean, it was heavy, but you couldn't carry it. [01:25] So did you imagine that you'd be here today in 2016?

[01:31] No, it was like we were sitting over there and there were [01:35] you know, 14 of us or something. And you're hacking on this new system? I mean, even that was like a, we were sitting around like looking at whiteboards, trying to talk about what we should do. Like this was a, [01:45] We're... [01:47] It's almost impossible to sort of... [01:49] overstate how much [01:53] we were like a research lab with no [01:55] With a very strong belief and direction and conviction, but no real kind of like action plan. [02:01] I mean, not only was like the idea of a company or a product sort of unimaginable. [02:06] the specific like LLMs as an idea we're still very far off. [02:10] And so... We're trying to play video games. Trying to play video games. Are you still trying to play video games? [02:16] Now we're pretty good at that. [02:20] All right. So it took you another six years for the first consumer product to come out, which is ChatGPT. Along the way, how did you sort of think about milestones that [02:30] To get something to that level. As like an accident of history, the first consumer product was not ChatGPT. That's right. It was Dolly. The first product was the API. [02:42] So we had built... [02:44] We kind of went through a few different things. [02:48] a few directions that we really wanted to bet on. Eventually, as I mentioned, we said, well, we've got to build a system to see if it's working. [02:54] And we're not just writing research papers. So we're going to see if we can play a video game well. [02:59] We're going to see if we can do a robot hand. We're going to see if we can do a few other things.

[03:02] And at some point in there, uh, [03:05] one person and then initially, and then eventually a team got excited about trying to do unsupervised learning and to build language models. And that led to GPT-1 and then GPT-2. And by the time the GPT-3 started, [03:17] We both thought we had something that was kind of cool, but we couldn't figure out what to do with it. [03:22] Um, and also we realized we needed a lot more money to keep scaling. [03:28] We had done GPT-3, we wanted to go to GPT-4. We're heading into the world of billion-dollar models. It's hard to do those as a pure science experiment, unless you're like a particle accelerator or something. Even then, it's hard. [03:39] we started thinking, okay, we both need to figure out how this can become a business that can sustain the investment that it requires. [03:47] like [03:48] we have a sense that this is heading towards something actually useful. And we had put GPT-2 out as model weights, and not that much had happened. One of the things that I had just observed about GPT-2 [04:02] companies' products in general is if you do an API, it usually works somehow on the upside. This was like true across many, many YC companies. And also that... [04:13] If you make something much easier to use, there's usually a huge benefit to that. So we're like, well, it's kind of hard to run these models that are getting big. We'll go write some software, do a really good job of... [04:24] running them and also will then rather than build a product because we couldn't figure out what to build um we will hope that somebody else finds something to build and so i forget exactly when but maybe it was like june of 2020 um

[04:36] we put out GPT-3 in the API. [04:39] And [04:41] The world didn't care, but sort of Silicon Valley did. They're like, oh, this is kind of cool. This is poignant at something. And there was this weird thing where like... [04:48] We got [04:49] Almost no attention from most of the world. [04:52] And some startup founders were like, oh, this is really cool. Or like, I mean, some of them are like, this is AGI. Um... [04:59] The only people that built... [05:01] real businesses with the GPT-3 API that I can remember were these company, a few companies that did like copywriting as a service. [05:09] That was kind of the only thing GPT-3 was over the economic threshold on. [05:15] But, [05:16] One thing we did notice, which eventually led to ChatGPT, is even though people couldn't build a lot of great businesses with the GPT-3 API, people love to talk to it in the playground. [05:27] And it was terrible at chat. We had not at that point figured out how to do our LHF to make it easy to chat with. But [05:32] People love to do it anyway. [05:34] And... [05:36] In some sense, that was the kind of only killer use other than copywriting of the API product. [05:42] that led us to eventually build ChatGPT. By the time ChatGPT 3.5 came out, there were maybe eight categories instead of one category, where you could build a business with API. [05:53] But... [05:55] But our conviction that people just want to talk to the model had gotten really strong. So we had done Dolly, and Dolly was doing okay, but we knew we kind of wanted to build... [06:04] especially along with the fine tuning we were able to do, we knew we wanted to build this model

[06:09] this product let you talk to the model. [06:10] And it launched in 2022 or something? Uh... [06:17] I think, yes. Yeah, about six years from when the first... November 30th, 2022. [06:21] Yeah. So there's a lot of work leading up to that and... [06:25] 2022 launched today. It has over 500 million people who talk to it on a weekly basis. [06:31] Yeah. All right. All right. So, by the way, get ready for some audience questions because that was Sam's request. You've been here for every single one of the Ascents, as Pat mentioned, and there's been lots of ups and downs. Yeah. [06:46] Seems like the last six months. [06:48] It's just been shipping, shipping, shipping. [06:50] shipped a lot of stuff, and it's amazing to see the product velocity, the shipping velocity... [06:56] continue to increase. So this is like multi sort of-- [07:00] Part question. How have you gotten a large company to like increase product velocity? [07:05] over time. I think a mistake that a lot of companies make is they get big and they don't do any, they don't do more things. So they just like get bigger because you're supposed to get bigger and they still ship the same amount of product. And that's when like... [07:18] the molasses really takes hold. I am a big believer that you want everyone to be busy. [07:23] You want teams to be small. You want to do... [07:27] a lot of things relative to the number of people you have. Otherwise, you just have like 40 people in every meeting and huge fights over who gets like what tiny part of the product. [07:36] There was this old observation of business that a good executive is a busy executive because you don't have people muddling around.

[07:45] But... [07:46] I think it's like a good... [07:50] At our company and many other companies, researchers, engineers, product people, they drive almost all the value. [07:56] And you want those people to be busy and high impact. So if you're going to grow, you better do a lot more things. Otherwise, you kind of just have... [08:04] a lot of people sitting in a room fighting or meeting or talking about whatever. So we try to have [08:09] You know, relatively small numbers of people with huge amounts of responsibility. And the way to make that work is to do a lot of things. [08:16] And also like... [08:18] We... [08:19] have to do a lot of things. I think we really do now have an opportunity. [08:26] to go build [08:28] one of these [08:29] important internet platforms. Um, [08:32] But to do that, like if we really are going to be [08:36] people's like [08:37] personalized AI that they use across many different services and, you know, over their life and across... [08:43] all of these different [08:45] all of these different main categories, and all the smaller ones that we need to figure out how to enable, [08:51] That's just a lot of stuff to go build. [08:54] Anything you're particularly proud of that you've launched in the last six months? [08:58] I mean, the models are so good now. They still have areas to... [09:03] get better, of course, and we're working on that fast, but like, [09:08] I think at this point, [09:10] ChatGPT is a very good product because the model is very good. I mean, there's other stuff that matters too, but

[09:15] The... [09:16] I am like... [09:17] I'm amazed that one model can do so many things so well. [09:20] Thank you. [09:21] You're building small models and large models. You're doing... [09:25] a lot of things, as you said. So how does this audience stay out of [09:30] Your ways and not be roadkill. [09:35] Um... [09:36] I mean, like, I think the way to model us is we want to build... [09:40] We want to be people's core AI subscription. [09:43] And... [09:44] way to use that thing. [09:46] Some of that will be like what you do inside of ChatGPT. [09:52] We'll have a couple of other kind of like really key... [09:55] parts of that subscription. [09:57] But mostly, [09:59] We will hopefully build this smarter and smarter model. We'll have these surfaces like... [10:05] future devices, future things that are sort of similar to operating systems, whatever. And then, you know, we want [10:13] We have not yet figured out [10:15] exactly I think what the sort of [10:17] API... [10:19] or SDK or whatever you want to call it is to like really be [10:22] our platform, but we will. It may take us a few tries, but we will. [10:26] And I hope [10:28] that that enables like just an unbelievable amount of wealth creation in the world and other people to build. [10:33] onto that but yeah we're gonna go for like the core ai subscription and the model [10:38] And then the kind of core surfaces. And there will be a ton of other stuff to build. [10:42] So don't be the core area of subscription, but you can do everything else.

[10:46] We're going to try. I mean, if you can make a better Core i subscription offering than us, go ahead. That'd be great. Okay. It's rumored that you're raising $40 billion or something like that at $340 billion valuation. It's rumored. [11:03] I don't know if that's true. I think we announced it. Okay, well, I just want to make sure that you announced it. What's your scale of ambition from here? [11:12] Thank you. [11:13] We're going to try to make great models and ship good products and [11:17] there's no master plan beyond that. Like we're going to, I think like, no, I, I, I mean, there's, I see plenty of opening eye people in the audience. They can vouch for that. Like we don't, [11:28] We don't sit there and have-- like, I am a big believer that you can kind of like-- [11:32] do the things in front of you, [11:34] But if you try to work backwards from like, we have this crazy complex thing. [11:38] Um, [11:40] that doesn't usually work as well. Like the, the, we, we know that we need tons of AI infrastructure. Like we know we need to go build out massive amounts of like, [11:50] AI factory volume. We know that we need to keep making models better. We know that we need to build a great top of the stack, [11:58] consumer product and all the pieces that go into that but [12:03] we pride ourselves on being nimble and adjusting tactics as the world adjusts. And so the product's... [12:10] Um [12:11] The products that we're going to build next year, we're probably not even thinking about right now. [12:15] and

[12:18] we believe we can build a set of products that people really, really love. Um, and we have like unwavering confidence in that. And we believe we, [12:26] can build great models. I've actually never felt more optimistic about our research roadmap than I do right now. What's on the research roadmap? Really smart models. But in terms of the steps-- [12:41] in front of us, we kind of take those one or two at a time. So you believe in working forwards, not necessarily working backwards? [12:48] I have heard... [12:49] Some people talk about these brilliant strategies of how this is where they're going to go and they're going to work backwards. And this is takeover the world. And this is the thing before that. And this is that. [12:59] And here's where we are today. I have never seen those people really massively succeed. [13:03] Got it. Who has a question? [13:07] Thank you. [13:08] There's a mic coming your way. Being thrown. [13:12] Thank you. [13:13] What do you think the larger companies are getting wrong about transforming their organizations to be more AI native in terms of both using the tooling as well as producing products? [13:25] you know, it's, [13:26] Smaller companies are clearly just beating the crap out of out of larger ones when it comes to innovation here. [13:31] Yeah. [13:32] I think this basically happens every... [13:35] major tech revolution. [13:38] There's nothing to me surprising about it. The thing that they're getting wrong is the same thing they always get wrong, which is like, [13:44] people get incredibly stuck in their ways, organizations get incredibly stuck in their ways. If things are changing a lot every quarter or two,

[13:51] And you have like [13:54] an information security console that meets once a year to decide what applications are going to allow and what it means to put data into a system like [14:01] It's just, it's so painful to watch what happens here. [14:05] You know, this is... [14:07] This is creative destruction. This is why startups win. This is how the industry moves forward. I am... [14:12] I'd say I feel disappointed but not surprised at the rate that big companies are willing to do this. They will... [14:22] My prediction would be that there's another couple of years of fighting pretending like this isn't [14:26] I'm going to reshape everything. [14:28] And then there's a capitulation and a last minute scramble. And it's sort of too late. [14:32] And [14:33] In general, startups just sort of like... [14:35] blow past people doing it the old way. [14:39] I mean, this happens to people, too. Like, watching... [14:42] Watching like a [14:45] Someone who started... [14:48] maybe you like... [14:50] talk to an average 20-year-old and watch how they use ChatGPT. [14:54] And then you go talk to like an average 35 year old and how they use it or some other service. And like the difference is unbelievable. It reminds me of like. [15:02] you know, when the smartphone came out and like every kid was able to use it super well and older people just like took like three years to figure out how to do basic stuff. And then, of course, people integrate. But but the [15:14] This sort of like generational divide on AI tools right now is crazy. And I think companies are just another symptom of that. [15:20] Anybody else have a question?

[15:23] Just to follow up on that, what are the cool use cases that you're seeing young people using with ChatGPT that might surprise us? [15:31] Thank you. [15:32] Thank you. [15:32] They really do use it like an operating system. They have complex ways to set it up, to connect it to a bunch of files, and they have fairly complex prompts memorized in their head, or in something where they paste in and out. [15:49] the [15:52] I mean, that stuff I think is all cool and impressive. And there's this other thing where like, they don't really make life decisions without asking like Chachi Biti, what they should do. And it has like the full context on every person in their life and what they've talked about. And you know, like the memory thing has been a real change there, but yeah, [16:06] But yeah, I think... [16:08] Gross oversimplification, but like, [16:12] Older people use ChatGPT as a Google replacement. [16:15] Maybe people in their 20s and 30s use it as like a... [16:19] life advisor something and then like [16:24] people in college use it as an operating system. [16:27] How do you use it inside of OpenAI? [16:31] I mean, it writes a lot of our code. How much? I don't know the number. And also, when people say the number, I think it's always this very dumb thing. Because like... Satya said Microsoft code is 20-30% written by... Measuring by lines of code is just such an insane way to... Like, I don't... [16:46] I would... [16:47] Maybe the thing I could say is it's writing meaningful code [16:51] I don't know how much, but it's like writing the parts that actually matter.

[16:56] That's interesting. Next question. [16:59] Oh. Mike going around? Is this-- OK. Hey, Sam. [17:02] I thought it was interesting that the... [17:07] answer to Alfred's question about where you guys want to go is focused mostly around consumer and being the core subscription and [17:12] and also most of your revenue comes from [17:15] consumer subscriptions, why keep the API in 10 years? [17:19] I really hope that all of this merges into one thing. [17:23] Like, you should be able to sign in with OpenAI to other services. Other services should have an incredible SDK to take over the ChatGPT. [17:31] Um... [17:32] UI at some point. [17:34] But to the degree that you are going to have [17:37] a personalized AI that knows you, that has your information, that knows what you want to share later and has all this context on you. [17:44] You'll want to be able to use that in a lot of places now. I agree that the current I [17:48] version of the API is very far off that vision, but I think we can get there. [17:54] Yeah, maybe I have a follow-up question to that one. You kind of took mine. But a lot of us who are building application layer companies, we want to use those building blocks, those different API components, maybe the deep research API, which is not a release thing, but could be, and build stuff with them. [18:12] Is that going to be a priority, like enabling that platform for us? How should we think about that? Yeah. [18:19] I think, I hope something in between those, that there is sort of like a new... [18:23] protocol on the level of HTTP.

[18:26] for [18:27] the future of the internet where things get federated and broken down into like much smaller components and agents are like constantly [18:35] exposing and using different tools and authentication, payment, data transfer. It's all built in at this level. [18:42] that everybody trusts, everything can talk to everything. [18:45] And... [18:48] E. [18:50] I. [18:51] I don't quite think we know what that looks like, but it's like coming out of the fog. [18:55] And as we get a better sense for that, again, it'll probably take us like a few... [19:00] iterations toward that to get there but that's kind of where i would like to see things go [19:06] Thanks. [19:08] Hey, Sam. Back here. My name's Roy. I'm curious. [19:13] Uh, [19:14] the AI would obviously do better with more input data. Is there any thought to feeding [19:20] sensor data and what type of sensor data, whether it's [19:25] temperature, things in the physical world that you could feed in that it could better understand reality. [19:32] People do that a lot. People like put that into... [19:35] People have whatever. They build things where they just put sensor data into an API, like an O3 API call or whatever. [19:41] And for some use cases, it does work super well. [19:44] I'd say that the latest models seem to do it [19:46] good job with this and they used to not. [19:49] We'll probably bake it in more explicitly at some point, but there's already a lot happening there. [19:54] Hi, Sam. I was really excited to play with the voice model in the playground. And so I have two questions. The first is how important is voice to open AI in terms of like stack ranking for infrastructure? And can you share a little bit about how you think it'll show up in the product and chat GPT, the core thing?

[20:14] I think voice is extremely important. Honestly, we have not made a good enough voice product yet. That's fine. It took us a while to make a good enough text model, too. [20:24] We will crack that code eventually. And when we do-- [20:28] I think a lot of people are going to want to use voice voice. [20:31] interaction a lot more i i am [20:35] Super... When we first launched... [20:37] our current voice mode the thing that was most interesting to me was it was a new stream on top of [20:43] like the touch interface. And you could talk and be like clicking around on your phone at the same time. [20:48] And I continue to think there's something amazing to do about like voice plus GUI interaction that we have not cracked yet. [20:56] But. [20:57] Before that, we'll just make voice really great. And when we do, I think there's not only... [21:02] Is it cool with existing devices? But I sort of think voice will enable... [21:06] a totally new class of devices if you can make it feel like truly human level voice. [21:12] Similar question. [21:13] I have a question about coding. I'm curious, is coding just another vertical application, or is it more central to the future of OpenAI? [21:19] That one's more central to the future of OpenAI. Coding, I think, will be how [21:25] these models... [21:28] kind of [21:31] Right now, if you ask ChatGPT a response, you get text back, maybe you get an image. You would like to get a whole program back. You would like custom rendered code for every response, or at least I would. You would like the ability for these models to go make things happen in the world.

[21:47] And writing code, I think, will be very central to how you actuate the world and call a bunch of APIs or whatever. So I would say coding will be more in... [21:58] a central category, we'll obviously expose it through our API and our platform as well. [22:02] But, [22:03] you know, chat GPT should be [22:05] excellent at writing code. [22:07] So we're going to move from the world of assistance to agents to basically applications all the way through? [22:13] I think it'll feel... [22:16] Yeah, it's like very continuous, but yes. [22:19] Thank you. [22:20] So you have conviction in the roadmap about smarter models. Awesome. I have this mental model. There's some ingredients like-- [22:27] More data, bigger data centers, a transformer architecture, test time compute. What's like an underrated ingredient or something that's going to be part of that mix that like maybe isn't in the mental model of most of us? [22:43] Thank you. [22:45] Um, [22:51] I mean, that's kind of the... Each of those things are really hard. And, you know, obviously, like... [22:56] Thank you. [22:58] The highest leverage thing is still big algorithmic breakthroughs. And I think there still probably are some 10x or 100x left. Not very many, but... [23:06] Even one or two is a big deal. Um... [23:10] but [23:10] It's like algorithms, data, [23:14] compute. [23:15] Those are sort of the big ingredients. Uh, hey.

[23:20] So my question is, you run one of the best ML teams in the world. How do you balance between letting smart people like ESA chase deep research or something else that seems exciting, versus going top down and being like, we're going to build this, we're going to make it happen, we don't know if it'll work? [23:39] There are some projects that require so much coordination that there has to be a little bit of like top down quarterbacking. [23:45] But I think most people try to do way too much of that. [23:52] I mean, this is like [23:54] There's probably other ways to run good AI research. [23:57] or good research labs in general. But when we started OpenAI, we spent a lot of time trying to understand... [24:04] Uh, [24:06] what a well-run research lab looks like and [24:10] You had to go really far back in the past. In fact, almost everyone that could help advise us on this was dead. [24:15] It had been like a long time since there had been good research labs. [24:19] And [24:22] You know, people ask us a lot, like, why... [24:25] Why does open AI like repeatedly innovate and why do the other AI labs like sort of copy or why do like, [24:31] BioLab X not do good work and BioLab Y does do good work or whatever. And, [24:36] We sort of keep saying, like, here's the principles we've observed. Here's how we learned them. Here's what we looked at in the past. [24:41] And then everybody says, great, but I'm going to go do the other thing. We said, that's fine. You came to us for advice. Do what you want. [24:49] But I find it remarkable how much these few principles that we've tried to run our research lab on, which we did not invent, we shamelessly copied from other good research labs in history, have worked for us. And then people who have had some smart reason about why they were going to do something else that didn't work.

[25:07] um so it seems to me that uh these large models uh one of the really fascinating things as like a lover of knowledge about them is that they potentially embody and allow us to answer these like amazing long-standing questions in the humanities about cyclical changes in artistic uh interesting things or even like uh [25:29] to what extent systematic prejudice and other sorts of things are really happening in society, and can we sort of detect these very subtle things which we could never really... [25:39] do more than hypothesize before. And I'm wondering whether OpenAI has a thought about or even a roadmap for working with academic researchers, say, to help unlock some of these new things we could [25:51] learn for the first time in the humanities and in the social sciences. [25:56] We do. [25:57] Yeah, I mean, it's amazing to see what people are doing there. We do have academic research programs where we partner and, you know, do some custom work. But mostly people just say, like, I want access to the model. Or maybe I want access to the base model. [26:09] And I think we're really good at that. One of the kind of... [26:14] cool things about what we do is so much of our incentive structure is pushed towards making the models as smart and cheap and widely accessible as possible. [26:22] that that serves academics and really the whole world very well. [26:25] So, [26:26] You know, we do some custom partnerships, but we often find [26:31] the [26:32] what [26:33] researchers or users really want is just for us to make the general model better across the board. So we try to focus 90 percent of our thrust vector on that.

[26:44] I'm curious how you're thinking about customization. So you mentioned the federated like sign in with OpenAI bringing your memories, your context. [26:51] I'm just curious if you think customization and these different post training on application specific things is a band-aid? [26:57] or trying to make the core models better and how you're thinking about that. [27:02] I mean, in some sense, I think the, like, platonic ideal state is, uh... [27:09] a very tiny reasoning model [27:12] with a trillion tokens of context. [27:13] that you put your whole life [27:15] into the model never retrain the weights never customized but that thing can like reason across your whole context and do it [27:21] efficiently. [27:22] And [27:23] every conversation you've ever had in your life, every book you've ever read, every email you've ever read, everything you've ever looked at is in there, plus connected all your data from other sources. [27:34] your life just keeps appending to the context, and your company just does the same thing for all your company's data. We can't get there today, [27:43] I [27:44] But. [27:45] But I think of kind of like anything else as a compromise off that platonic ideal. [27:50] And [27:51] That is how I would eventually... [27:53] I hope we do customization. One last question in the back. Hi, Sam. Thanks for your time. Where do you think most of the value creation will come from in the next 12 months? Would it be maybe... [28:04] advanced memory capabilities [28:07] or maybe security... [28:08] or protocols that allow agents to do more stuff and interact with the real world?

[28:14] Thank you. [28:15] Thank you. [28:16] Um... [28:17] Thank you. [28:20] Thank you. [28:21] I mean, in some sense, the value will continue to come from... [28:25] really three things like building out more infrastructure, smarter models, [28:29] And [28:30] building the kind of scaffolding to integrate this stuff into society. And if you push on those, I think the rest will sort itself out. [28:37] At a higher level of detail, [28:41] I kind of think 2025 will be [28:43] a year of sort of [28:45] agents doing work. Coding in particular, I would expect to be a dominant category. [28:50] I think there'll be a few others too. [28:52] Um, [28:53] Next year is a year where I would expect more like a... [28:57] sort of [28:58] AI is discovering new stuff and maybe we... [29:00] have AIs make some very large scientific discoveries or assist humans in doing that. And, you know, I'm, I am kind of a, [29:07] believer that [29:08] most of the sort of real sustainable, um, [29:11] economic growth in human history comes from, once you've like kind of spread out and colonized the earth, most of it comes from just [29:18] better scientific knowledge and [29:20] and then implementing that for the world. And then 27, I would guess, is the year where like, [29:25] that all moves from the sort of intellectual realm to the physical world and, [29:28] robots go from a curiosity to like a serious economic creator of value. [29:34] But that was like an off the top of my head kind of guess right now. [29:37] Can I close with... [29:39] A few quick [29:40] questions, one of which is GPT-5. Is that going to be just all smarter than all of us here?

[29:49] Um... [29:50] I mean, if you think you're like way smarter than O3, then maybe you have a little bit of a ways to go. But O3 is already pretty smart. [29:58] Yeah. [29:59] Two personal questions. Last time you were here, you had just come off a blip. [30:04] with [30:05] OpenAI. [30:06] Given some perspective now and distance, you got any advice for founders here about resilience, endurance, resilience? [30:15] strength, [30:16] Thank you. [30:19] Um, [30:21] Thank you. [30:23] It gets easier over time. I think like... [30:26] you will face a lot of [30:28] adversity in your journey as a founder and [30:32] the kind of challenges get... [30:34] harder and higher stakes, but [30:37] the emotional toll [30:39] gets easier as you kind of go through more bad things. So it's, uh, [30:43] you know, in some sense, like... [30:46] Thank you. [30:46] It does, it does, yeah, even though like, [30:49] abstractly the challenges get bigger and harder, [30:53] your ability to deal with them, the sort of resilience you build up gets easier, like with each one. [30:57] you kind of go through. Um... [31:00] Thank you. [31:03] And then I think the hardest thing about... [31:07] the big challenges that come, [31:09] as a founder is not the moment when they happen. Like a lot of things go wrong in the history of a company. [31:16] Um, [31:17] in the acute thing, you can kind of like

[31:21] You get a lot of support, you can function a lot of adrenaline, [31:26] even the really big stuff like your company runs out of money and fails like a lot of people will come and support you um and you kind of get through it and go on to the new thing the thing that [31:35] I think it's harder to sort of manage your own psychology through. [31:39] is the sort of like fallout after. [31:41] um and i think if there's [31:44] People focus a lot about how to work in that one moment during the crisis, and the really valuable thing to learn is, [31:51] is how you pick up the pieces. [31:54] There's much less talk about that. [31:55] I think there's [31:56] I've never actually found something good to point [31:59] founders to go read about not how you deal with the real crisis on day zero or day one or day two, [32:05] But on day 60, as you're just trying to like... [32:08] rebuild after it. [32:10] And that's the area that I think you can practice and get better at. [32:14] Thank you, Sam. You're officially still on paternity leave. I know. So thank you for coming in and speaking with us. Appreciate it. Thank you.