The Exchanges

Q Do you think so? And you think they will stay in the UK?

A Let's say you really believed that you could build Let's say you believed you could build, as I framed it, and you quoted, like, an anthropic or open AI scale company in terms of ambition. I actually think one of your biggest challenges building that in the Bay Area today would be building and retaining the AI research talent that makes that, like, and, you know, people always talk about hiring. People much less often talk about retention. But, like, the packages that Sam Altman will offer to your best people the minute you get your first, like, Uh, you know, demo. You know, it's really hard. I think in, in London, it's much more plausible that you build a truly world-class research team and keep them together for long enough to really see the impact. And you know, like, here in London, we have DeepMind, we have the London office of, we have the European office of Anthropic, the European office of OpenAI, we have WAVE, we have, you know, Oxford, Cambridge, Imperial, UCL, like, we, like, the, the bench is very deep when it comes to AI talent.

AI assessment note: “in London, it's much more plausible that you build a truly world-class research team”

Q It is never when you're 25 and you come out of Oxford and you're like, oh, this entrepreneurship thing. Yeah. Do you disagree with me having seen so many different entrepreneurial paths that you've seen?

A We believe one of our kind of cool ways that we evaluate founders entrepreneur first is that the best predictor of future behavior is past behavior. Um, so we absolutely think that it's very unlikely that you rock up with a perfect CV, but like no sign whatsoever Have ever having done anything that you weren't told to do by a teacher or a boss, and suddenly you're going to figure out how to do things that you aren't told to do. So in that sense, I think I would very strongly agree. I guess what I would say, though, is that, um, culture and default paths really matter, and so one thing we've learned about evaluating talent is what it looks like to do something that your parents, teachers, bosses didn't tell you to do in the Bay Area Is very different from in Singapore. You know, like, if you grew up in the Bay Area, and you go to Stanford, and you graduate, and you never thought about starting a company, and there's a very negative signal on you as a founder. If you grew up in Palani in, you know, rural India. Um, I don't think I need to have seen that you started a company before, but what I'm doing when I'm interviewing you is figuring out what is the behavior? What is your equivalent of the story I just told you? What is your equivalent of that entrepreneurial activity that is a predictor of you being able to succeed in an unstructured environment? That's what I care about.

AI assessment note: “So in that sense, I think I would very strongly agree.”

Q Do you think so? And you think they will stay in the UK?

A Let's say you really believed that you could build Let's say you believed you could build, as I framed it, and you quoted, like, an anthropic or open AI scale company in terms of ambition. I actually think one of your biggest challenges building that in the Bay Area today would be building and retaining the AI research talent that makes that, like, and, you know, people always talk about hiring. People much less often talk about retention. But, like, the packages that Sam Altman will offer to your best people the minute you get your first, like, Uh, you know, demo. You know, it's really hard. I think in, in London, it's much more plausible that you build a truly world-class research team and keep them together for long enough to really see the impact. And you know, like, here in London, we have DeepMind, we have the London office of, we have the European office of Anthropic, the European office of OpenAI, we have WAVE, we have, you know, Oxford, Cambridge, Imperial, UCL, like, we, like, the, the bench is very deep when it comes to AI talent.

AI assessment note: “I think in, in London, it's much more plausible that you build”

Q I'm really intrigued that we, we spoke about ideas and this isn't in the schedule, but I'm, I'm fascinated by it because obviously as we said with the FP in entrepreneurs and amazing individuals come without ideas in some cases, how do you guys approach the idea creation phase?

A When we say that we'll take people pre idea, we don't mean that they shouldn't come with ideas or even that they don't come with ideas. What we mean is that the basis of evaluation, the reason that we're selecting people is not, not the ideas that they come with. So we're not saying, you know, that sounds, that sounds like a terrible idea. That sounds like a good idea. Where the ideas come from at EF is an idea that we, uh, call edge. And what we mean by edge is what is a given person's, uh, competitive advantage that comes out of their skills or knowledge or experience Which gives them the potential to be globally competitive in, in some way. So edge, ah, is, you know, kind of key to everything we do right through from how we find and select people for the program right through to how they develop ideas and ultimately to how they, you know, raise funding at the end. And what we try and do during the process of, of selection is to make sure that our cohort is full of people who have really clear Competitive edges in one or more areas. Now often that's technical. They may be, um, uh, you know, they may have a very, very clear, uh, set of abilities in a particular, uh, technical field. Sometimes it's about a domain or a problem where they kind of may have some specialist knowledge. Where the ideas themselves come from, how you go from edge to idea, is that effectively you have a …

AI assessment note: “Where the ideas themselves come from, how you go from edge to idea”

Q Why do you like writing immersive murder mystery games?

A Um, I think... That most of us have this itch that is very hard to scratch, which is to be in these completely alien environments where we get to be someone else and do something else, you know, in a way where no one is in charge. I, you know, if you look at things like secret cinema and how popular that's been, I kind of feel like it's really interesting. People love dressing up and, you know, doing the thing, but like, it kind of, like, you dress up, you watch the film. I think most people have an itch to see, like, what would it be like to actually have to, like, live through this scenario, but with no consequences? And so I've, I've, I've written a series of historical murder mystery games where you don't know, it's not predetermined who or if anyone will be killed. You, there's 12 characters, you each have a set of goals and relationships, And the evening just unfolds, and you decide what happens. And I love playing them. I love writing them. I think it's a sort of human itch that there are very few ways to scratch, and it's like some of the most fun that I think you can have. You can see them on my website, Harry, for your next dinner party.

AI assessment note: “I love playing them. I love writing them. I think it's a sort of human itch”

Q It is never when you're 25 and you come out of Oxford and you're like, oh, this entrepreneurship thing. Yeah. Do you disagree with me having seen so many different entrepreneurial paths that you've seen?

A We believe one of our kind of cool ways that we evaluate founders entrepreneur first is that the best predictor of future behavior is past behavior. Um, so we absolutely think that it's very unlikely that you rock up with a perfect CV, but like no sign whatsoever Have ever having done anything that you weren't told to do by a teacher or a boss, and suddenly you're going to figure out how to do things that you aren't told to do. So in that sense, I think I would very strongly agree. I guess what I would say, though, is that, um, culture and default paths really matter, and so one thing we've learned about evaluating talent is what it looks like to do something that your parents, teachers, bosses didn't tell you to do in the Bay Area Is very different from in Singapore. You know, like, if you grew up in the Bay Area, and you go to Stanford, and you graduate, and you never thought about starting a company, and there's a very negative signal on you as a founder. If you grew up in Palani in, you know, rural India. Um, I don't think I need to have seen that you started a company before, but what I'm doing when I'm interviewing you is figuring out what is the behavior? What is your equivalent of the story I just told you? What is your equivalent of that entrepreneurial activity that is a predictor of you being able to succeed in an unstructured environment? That's what I care about.

AI assessment note: “I think I would very strongly agree.”

Q Actually, we've seen, you know, Incredible amounts of compute, ah, but it's data that is the bottleneck. Do you agree? And if you were to choose one of the three that is the bottleneck to the progression, what would it be?

A So I think it's certainly true that, um, Data is the one of those that is least obvious how you continue to, to scale. You know, as in, at the moment you can just translate dollars into compute relatively easy, easily, and so, you know, you can imagine at least a couple of orders of magnitude of, of growth of compute, you know, quite straightforwardly. It's not obvious that you can do the same in data. That said, I think people probably underestimate What we haven't done yet. I mean, like, we are not, you know, partly because of compute limits, you know, like, these models are not trained on all video, you know, in fact, most, you know, mostly that hasn't happened yet, and it's kind of interesting, I mean, some of the things that, um, people like, uh, Andre Carpathie have been talking about recently is, like, you know, the potential of, of thinking about video as being a, a really great way for a, um, For a model to build a world model and, you know, to understand more about how the world works. So yeah, data is a bottleneck, but I wouldn't bet against smart people figuring out ways to either create or ingest new types of data. I guess what I'm really arguing when I say we're on an S-curve is I don't think there's a lot further to go in just finding more text. You know, I think, I think we may be at the flattening out of the S-curve on text, but I suspect the next S-curve could…

AI assessment note: “So yeah, data is a bottleneck, but I wouldn't bet against smart people”

Q Why is that? I thought they would be all out dominant. We need to win this race and we will do anything to win it.

A I think you've just got to see the CCP as being, like, fundamentally extra, have an extreme priority on stability. You know, like the, the whole story of the CCP and its, and its dominance is eliminate threats to stability. And, you know, whatever you think about AI, whether you're bullish or bearish, whether you care about safety, whether you don't care about safety, I think everyone agrees that it's like a destabilizing force. And so I think the last thing they want is either companies gaining a lot of power, uh, through AI or AI itself being destabilizing. So, you know, like if you read the, the, the regulation that's already coming through, For example, if you train a large model, you have to, like, supply random samples of the training data to the, um, to the CCP or to the government. You have to, uh, you have to, like, show it's, like, political, that it, that it is, um, not gonna undermine any of the, you know, state positions of the CCP on various issues. So, like, it's actually a very, very regulated environment. Now, that doesn't mean that ultimately the Chinese state won't try and harness powerful AI for its own ends, as you're sort of talking about. But I think it's easy to Understate the level of paranoia they have about it. So, you know, my, my view would be, there, there are two factors that made me think that maybe, you know, maybe I'm not quite in Alex's camp. …

AI assessment note: “the whole story of the CCP and its, and its dominance is eliminate threats to stability”

Q Do you think we're gonna hit that flattening again?

A Well, I think it's a race, as in like, I do think that if, if we can't find anything better than just trying to scale LLMs, then yeah, the flattening out will be brutal. But my, if I had to bet, my bet is that the enormous amount of capital and talent That has been aggregated, you know, around this relatively small number of companies because of the hype will be enough to unlock the next S-curve, and so you'll see, you know, kind of further improvement. If we can't, but this is what I mean about the value, the returns to good ideas, I think will be very high over the next few years. So actually another way of thinking about it is one thing you've seen over the last couple of years since GPT-IV was released is actually a convergence of capabilities. You know, when GPT-IV came out, it was very clear that OpenAI were In the lead. And now, like, there's quite a few companies that have a GPT-IV-ish level model. You see convergence, right? My guess is that you may see more divergence over the next few years, because the value of ideas goes up. Because people aren't just scaling.

AI assessment note: “my bet is that the enormous amount of capital and talent”

Q I'm gonna read the book. Uh, so my question to you on, and actually this was, I spoke to Charlie Songhurst last night, and he actually gave me this one. He said, how much of a great founding team is the individual ability of the founders versus the synergies between those founders?

A This is something I've changed my mind on. Um, I actually think more and more that it's really about the synergies matter, but ultimately I do think entrepreneurship is one of the paths that most relies on the peak performance of the highest performer. So there are some walks of life where what really matters is average performance. You know, like, how well do you do, you know, on a, uh, on the median day? And there are other walks of life where it's like, how well do you do at your very best? I think you need people in every company who are in the former category, who can day in, day out, turn out excellence. But I increasingly believe Let's use Helsing as an example you already brought up. Like, you also just need the peak performance of the best day from the top talent in the company to be truly exceptional. And so like, yeah, synergies do matter, but I think probably the best predictor is how good is the best person at their best.

AI assessment note: “synergies do matter, but I think probably the best predictor is how good is the best person”

Q Actually, we've seen, you know, Incredible amounts of compute, ah, but it's data that is the bottleneck. Do you agree? And if you were to choose one of the three that is the bottleneck to the progression, what would it be?

A So I think it's certainly true that, um, Data is the one of those that is least obvious how you continue to, to scale. You know, as in, at the moment you can just translate dollars into compute relatively easy, easily, and so, you know, you can imagine at least a couple of orders of magnitude of, of growth of compute, you know, quite straightforwardly. It's not obvious that you can do the same in data. That said, I think people probably underestimate What we haven't done yet. I mean, like, we are not, you know, partly because of compute limits, you know, like, these models are not trained on all video, you know, in fact, most, you know, mostly that hasn't happened yet, and it's kind of interesting, I mean, some of the things that, um, people like, uh, Andre Carpathie have been talking about recently is, like, you know, the potential of, of thinking about video as being a, a really great way for a, um, For a model to build a world model and, you know, to understand more about how the world works. So yeah, data is a bottleneck, but I wouldn't bet against smart people figuring out ways to either create or ingest new types of data. I guess what I'm really arguing when I say we're on an S-curve is I don't think there's a lot further to go in just finding more text. You know, I think, I think we may be at the flattening out of the S-curve on text, but I suspect the next S-curve could…

AI assessment note: “So I think it's certainly true that, um, Data is the one of those”

Q Do you think we're gonna hit that flattening again?

A Well, I think it's a race, as in like, I do think that if, if we can't find anything better than just trying to scale LLMs, then yeah, the flattening out will be brutal. But my, if I had to bet, my bet is that the enormous amount of capital and talent That has been aggregated, you know, around this relatively small number of companies because of the hype will be enough to unlock the next S-curve, and so you'll see, you know, kind of further improvement. If we can't, but this is what I mean about the value, the returns to good ideas, I think will be very high over the next few years. So actually another way of thinking about it is one thing you've seen over the last couple of years since GPT-IV was released is actually a convergence of capabilities. You know, when GPT-IV came out, it was very clear that OpenAI were In the lead. And now, like, there's quite a few companies that have a GPT-IV-ish level model. You see convergence, right? My guess is that you may see more divergence over the next few years, because the value of ideas goes up. Because people aren't just scaling.

AI assessment note: “if I had to bet, my bet is that the enormous amount of capital”

Q But that's just really interesting, because everyone is saying, ah, we're seeing the complete commoditization. We even have Chinese players like Yi, I think it's called. Yeah. Which is like, almost equivalent of GPT-IV's performance. Are we not seeing the conversation?

A I think the pure LLM approach absolutely is commoditized. Will be, either is or will be very quickly commoditized. I guess what I'm saying is, I don't think GPT-V is just going to be a bigger language model. I think it's going to involve something different. It's going to involve kind of, uh, you know, like a lot of productizations, maybe search as we already, you know, discussed and things. And so like, but these are not commodity ideas yet. It's not yet, ah, the case that everyone will then just say, ah, or, or that immediately everyone will be able to copy that. And so I think, you know, broadly, you know, I was talking recently to someone who was on the Llama-III team, and they were like, you're probably underring the extent to which Llama-III is literally just Llama-II with more compute and more data. There's not a lot that's different other than, you know, that. I guess what I'm saying is Llama-IV will not just be You know, a bigger Lama three. It will have to be different. It will incorporate new ideas.

AI assessment note: “I think the pure LLM approach absolutely is commoditized.”

Q Why is that? I thought they would be all out dominant. We need to win this race and we will do anything to win it.

A I think you've just got to see the CCP as being, like, fundamentally extra, have an extreme priority on stability. You know, like the, the whole story of the CCP and its, and its dominance is eliminate threats to stability. And, you know, whatever you think about AI, whether you're bullish or bearish, whether you care about safety, whether you don't care about safety, I think everyone agrees that it's like a destabilizing force. And so I think the last thing they want is either companies gaining a lot of power, uh, through AI or AI itself being destabilizing. So, you know, like if you read the, the, the regulation that's already coming through, For example, if you train a large model, you have to, like, supply random samples of the training data to the, um, to the CCP or to the government. You have to, uh, you have to, like, show it's, like, political, that it, that it is, um, not gonna undermine any of the, you know, state positions of the CCP on various issues. So, like, it's actually a very, very regulated environment. Now, that doesn't mean that ultimately the Chinese state won't try and harness powerful AI for its own ends, as you're sort of talking about. But I think it's easy to Understate the level of paranoia they have about it. So, you know, my, my view would be, there, there are two factors that made me think that maybe, you know, maybe I'm not quite in Alex's camp. …

AI assessment note: “have an extreme priority on stability. You know, like the, the whole story”

Q There you go. I'm really worried about, like, knowledge inequality, actually. And I think we sit here, and we know all these things. Dude, like, a million people in the UK of sixty-five million have any idea what we're talking about? Maybe less. Do you not think this is going to create an ever-increasing chasm in wealth inequality?

A I think one of the good things is people benefit from technology even when they don't understand it. I think that's like one of the big lessons in, you know, the history of the last 200 years of capitalism is that, um, uh, you know, you, technology is, is the engine of prosperity and, you know, like people whose jobs involve very little technology today make a whole lot more than their ancestors 200 years ago and have a much better quality of life because of the benefits that technology brings. Um, I do think it's true that so far, you know, and the broad trend is that, uh, Technology is what I guess economists call skills biased. It benefits people with more skills more than it benefits people with, with fewer skills. Um, and that's a big challenge. That is a big challenge for governments. It's probably a big opportunity for, for entrepreneurs. But I think in general, the story of technology is, is an extraordinarily positive one. And I think, although I, I'm not saying that people shouldn't, including policymakers, look at this and say, right, how do we plan for this world? Where relatively few people understand this, but it's a huge factor in everyone's lives. But I worry about that, because I worry that you start to get the instinct of like, maybe we should try and slow it down, maybe we should try and stop it, and...

AI assessment note: “people benefit from technology even when they don't understand it”

Q What about people who say, ah, but the liquidity markets are shit. Like LSE is useless. There's not enough of a buy book. Big institutions aren't understanding enough. How do you think about that?

A I think they're right, but like, well, you know this, like, This is a, the way I see is, this is really a question about who benefits. It's not actually a question about whether we can build. What I mean by that is, I would love it that when, you know, a 20 VC company or an EF company goes public, that the, you know, the pools of capital that own that company are my mom and dad's, you know, West Yorkshire pension fund. Um, right now, that's not what's gonna happen, because they're gonna IPO in places that Uh, the shareholders are probably American, you know, so, Vanguard and BlackRock, right? And honestly, for you and me, and for the entrepreneurs, Who cares? You know, like, it's, um, it's, it's all cash. But again, if you want to, you know, from a UK, like, long-term growth perspective, I believe that so much of our growth, if we make it happen, will come from, you know, like, entrepreneurship, like, big advances in, in, in science and technology. And I think, frankly, capital markets are very efficient. Like, I think good things will raise capital. The question is, who's capital? And I think, I think from a UK PLC perspective, we should want it to be that it's pension funds You know, with UK pensions, UK savers that are benefiting from that. So that's a big problem.

AI assessment note: “I think they're right, but like, well, you know this, like, This is a”

Q I'm gonna read the book. Uh, so my question to you on, and actually this was, I spoke to Charlie Songhurst last night, and he actually gave me this one. He said, how much of a great founding team is the individual ability of the founders versus the synergies between those founders?

A This is something I've changed my mind on. Um, I actually think more and more that it's really about the synergies matter, but ultimately I do think entrepreneurship is one of the paths that most relies on the peak performance of the highest performer. So there are some walks of life where what really matters is average performance. You know, like, how well do you do, you know, on a, uh, on the median day? And there are other walks of life where it's like, how well do you do at your very best? I think you need people in every company who are in the former category, who can day in, day out, turn out excellence. But I increasingly believe Let's use Helsing as an example you already brought up. Like, you also just need the peak performance of the best day from the top talent in the company to be truly exceptional. And so like, yeah, synergies do matter, but I think probably the best predictor is how good is the best person at their best.

AI assessment note: “synergies do matter, but I think probably the best predictor is how good is the best person”

Q What did EF get most wrong in your assumptions on talent?

A You know, one of the challenges of this business, and you may feel the same, is that the feedback loops are so long. And so, you know, you, you do something, you observe a short-term metric, like, do they raise the seed round? Do they raise the series A? And then years later, you find out whether the company's any good. And so I think one of the things we got wrong was we just overreacted to short-term data early on. So, you know, we, in the first couple of years, we, Largely funded people who were straight out of university, and, um, a lot of them did very well. And then as the brand grew and the track record grew, we got more and more applicants who were, like, in their sort of, like, thirties. We're like, wow, this is great. Like, we're, we're able to, like, move up the, um, the, the experience curve. And of course, the problem was they were more experienced But they weren't the very best thirty-year-olds, whereas we could actually get the very best thirty-one-year-olds. You know, like, sorry, twenty-one-year-olds. Um, yeah, so as in, we, one of the things that I think we made a mistake on is that as we scaled, and particularly as Alice and I stopped making every selection decision, you got into the thing where it's very hard if you're new to what we do to compare the CV of like a thirty-one-year-old Who's actually good but not exceptional, and a 21 year old who's exceptiona…

AI assessment note: “one of the things we got wrong was we just overreacted to short-term data”

Q I'm really intrigued that we, we spoke about ideas and this isn't in the schedule, but I'm, I'm fascinated by it because obviously as we said with the FP in entrepreneurs and amazing individuals come without ideas in some cases, how do you guys approach the idea creation phase?

A When we say that we'll take people pre idea, we don't mean that they shouldn't come with ideas or even that they don't come with ideas. What we mean is that the basis of evaluation, the reason that we're selecting people is not, not the ideas that they come with. So we're not saying, you know, that sounds, that sounds like a terrible idea. That sounds like a good idea. Where the ideas come from at EF is an idea that we, uh, call edge. And what we mean by edge is what is a given person's, uh, competitive advantage that comes out of their skills or knowledge or experience Which gives them the potential to be globally competitive in, in some way. So edge, ah, is, you know, kind of key to everything we do right through from how we find and select people for the program right through to how they develop ideas and ultimately to how they, you know, raise funding at the end. And what we try and do during the process of, of selection is to make sure that our cohort is full of people who have really clear Competitive edges in one or more areas. Now often that's technical. They may be, um, uh, you know, they may have a very, very clear, uh, set of abilities in a particular, uh, technical field. Sometimes it's about a domain or a problem where they kind of may have some specialist knowledge. Where the ideas themselves come from, how you go from edge to idea, is that effectively you have a …

AI assessment note: “how you go from edge to idea, is that effectively you have a hundred extraordinarily smart”

Q So talk to me, Matt. We've got Cambridge, we've got MIT, McKinsey. It would have been very easy for you to continue down this very traditional and very successful path. So what caused the pivotal shift to founding Entrepreneur First, and what was that aha moment for you?

A Well, actually, it's funny that you should, like, kind of list those, uh, credentials or badges, if you like, in the, uh, my most popular, I should probably say only popular, a blog post is a blog post called Don't Be a Badge Collector. It's written to aspiring founders, but it really does apply, I suppose, to my own, and I think to Alice's story as well, in the, uh, I suppose what we're saying and what we believe founding thesis of EF is that The way that, uh, European society has evolved is such that the most ambitious people in our society almost see their careers as a succession of more badges to collect, you know, tick off doing well at schools, that you can tick off doing well at universities, that you can tick off getting a great first job, that you can tick off getting a promotion, et cetera, et cetera, et cetera. And I think from relatively early on in, in, in my kind of career, I felt that there had to be something else and that actually badge collecting, um, you know, satisfying as it was and, you know, kind of as twinkly as it made your CV, it wasn't really doing enough. And so for, for me, uh, starting entrepreneur first was really about how do I have impact? How do I kind of scale my impact beyond what would be possible through a conventional career?

AI assessment note: “starting entrepreneur first was really about how do I have impact?”

Q And I want to finish then on the future of EF this before we dive into a quick fire and you're expanding into Singapore. So I'm intrigued. What's it about one location like Singapore that allows it to be right for disruption with EF over maybe another?

A Yeah, sure. So, um, so I've already said that, you know, I think you need, you need a sense of mission to be a founder. And, you know, obviously EF is, is very much a kind of mission driven company too. And our mission is goes back to that very first thing we talked about, which is about What do careers look like in this century? And what does ambition look like in this century? So our kind of deep belief is that Entrepreneur First should work wherever there is a pool of talent that is, uh, really extraordinary, both in its, uh, technical ability and its ambition. Uh, where there is an emerging startup ecosystem, uh, that, that means that there's the kind of support and, and capital around it, uh, to allow the ambition to flourish. And third, where there's a supportive business environment in which, ah, you know, it's, it's possible to build and scale companies. You know, for us, Singapore ticked those boxes. You know, you have extraordinary technical talent, both within Singapore and within the kind of broader region. Two, you have an emerging VC and, and startup ecosystem that's really just starting to take off, but it's still early enough that it's not played out. And third, it's, you know, a very easy place to do business. So when we thought about Singapore, it ticked the boxes. You know, what was exciting for us is that it did a better job of that than, than some places th…

AI assessment note: “You know, for us, Singapore ticked those boxes.”

Q And you mentioned entrepreneurship there, and then I have to ask, can entrepreneurialism be taught, or is it an innate skill within oneself?

A I mean, I, I don't think it's either. I mean, we, we've often talked at Entrepreneur First about, um, your, your abilities as a founder are something that you discover by being a founder. Uh, I don't think they're either latent within you at birth and you, you know, you just, ah, you are or you're not, nor do I believe that you can sit down in the classroom and be taught it. What I would say is that how strong your entrepreneurial skills are, how kind of strong your, um, abilities are as a founder is, is a kind of almost, I suppose in technical terms, it's procedurally generated. It happens as a result of the actions you take by being a founder. And, you know, so the way I encourage people to think about it is, It may well be that they're, they have what it takes to be a superb founder, but the only way they can find that out is by trying. Um, there is no test you can take. Um, there is no, um, you know, personality, psychographic, multiple choice quiz that will tell you whether you're a founder. Really the only way to find out is by trying because you get better by being one.

AI assessment note: “I don't think it's either. I mean, we've often talked at Entrepreneur First”

Q the skills and abilities there and how they kind of grow within you from founding. So I'm really intrigued then, because a lot of founders say to me, you know, I'd love to start a startup, but I don't think I have the skills. I'm just not ready. So I'm intrigued to what you think ready is, and, and how we know when one is ready. What are the signs?

A Yeah. So if you go back to this idea of badge collecting, the, uh, on the side of the badge collectors, the most, uh, you know, dangerous, uh, weapon or the most dangerous obstacle to actually getting started and being a founder is this idea of readiness. I think it's very, very dangerous and very, very negative for people to believe that they need to wait until they're, in inverted commas, ready before they start a startup. I don't believe that anyone is ready. Um, I think actually what founding a company throws at you is so unpredictable, um, And so challenging, and so vast in its variety, that actually being ready is, is almost impossible. Which is again, why I say that you discover your abilities. You know, what, what it means to be a great entrepreneur actually varies from, from startup to startup. And so I think waiting for readiness is, is the ultimate red herring. Uh, and in fact, if you go back and you look at the stories of the really great founders, you know, the, the, in one sense they were ready by definition in that they succeeded. But also if you look at who they were and what their background was and what their skills were, you'd have to say that they weren't qualified in any, uh, normal sense of the word. It just so happened that because of who they were intrinsically, they were able to become ready through, through that process of discovery.

AI assessment note: “I don't believe that anyone is ready... waiting for readiness is, is the ultimate red herring”

Q What did EF get most wrong in your assumptions on talent?

A You know, one of the challenges of this business, and you may feel the same, is that the feedback loops are so long. And so, you know, you, you do something, you observe a short-term metric, like, do they raise the seed round? Do they raise the series A? And then years later, you find out whether the company's any good. And so I think one of the things we got wrong was we just overreacted to short-term data early on. So, you know, we, in the first couple of years, we, Largely funded people who were straight out of university, and, um, a lot of them did very well. And then as the brand grew and the track record grew, we got more and more applicants who were, like, in their sort of, like, thirties. We're like, wow, this is great. Like, we're, we're able to, like, move up the, um, the, the experience curve. And of course, the problem was they were more experienced But they weren't the very best thirty-year-olds, whereas we could actually get the very best thirty-one-year-olds. You know, like, sorry, twenty-one-year-olds. Um, yeah, so as in, we, one of the things that I think we made a mistake on is that as we scaled, and particularly as Alice and I stopped making every selection decision, you got into the thing where it's very hard if you're new to what we do to compare the CV of like a thirty-one-year-old Who's actually good but not exceptional, and a 21 year old who's exceptiona…

AI assessment note: “one of the things we got wrong was we just overreacted to short-term data”

Q That's a cracking question, actually. I love that one. Uh, and then let's finish on this. I know what you're going to say here, but the most exciting EF company And why? Or the EF company you'd most like to have joined if you're an entrepreneur? And why? Just don't say they're like my kids.

A Well, it genuinely is hard to pick one because one of the real privileges about EF is that we will get to what is such a huge range of founders. I mean, I haven't really talked about the companies at all yet, but you know, founders have built, and we've helped them build, I think, 75 companies now, and you know, they're worth now hundreds of millions of dollars. They're doing things in every field from agriculture to manufacturing to philanthropy to healthcare to education, so it is hard to pick. I do have a soft spot for companies that are using machine learning in interesting ways, and just to pick ones simply from the most recent cohorts, they're not trying to pick an all-time favorite, you know, we have a company like Third Eye, which is using machine learning to automate the use of CCTV and detection of crime. We have Uh, company like Cloud NC, which is using machine learning to automate, um, uh, the way that, uh, computer numerically control manufacturing works, which could have, you know, radical impacts there. We've got a company like Avalon AI using machine learning to detect, uh, dementia early. So very, very, um, broad range of applications for something like machine learning, and I'm just so excited about, um, about the impact that could have in, in the years to come.

AI assessment note: “company like Third Eye, which is using machine learning to automate the use of CCTV”

Q Do you think the world is ready for an agentic capable model to be unleashed?

A Not one that, where, like, many millions of instances can be run, and, um, they can operate at, like, nearly human level. Now, I don't think that's what GPT-V will be, but, like, I, you know, one of the, one of the theses I'm really interested in right now is, I think we'll get really good agents in the next five years. And I think if you look at parts of the economy where we already have a lot of automation, it's pretty clear that that requires a lot of infrastructure, requires protocols for agents to interact. I think we're going to need that for AI agents. We're going to need, ah, to build great infrastructure to maximize the economic value of these agents.

AI assessment note: “Not one that, where, like, many millions of instances can be run”

Q to founders because it's such a big topic that I do want to kind of really touch on properly, but I want to start on this tweet of yours, which I just thought was fantastic. And you said two predictions. One, someone will build a new open AI, anthropic scale AI company that takes the commoditization of LLMs as one of its core premises. What did you mean by this?

A I was very lucky to start investing in AI about 10 years ago, a bit more than that now, when the first sort of deep learning, uh, wave kicked off. And, you know, what's really interesting is that really, Certainly the last five years, but arguably longer, the story of AI is really a story largely about the deployment of just enormous amounts of compute and enormous amounts of data, and not that many new ideas, candidly. You know, I mean, like, I'm not saying there are none, but, like, broadly, if you had to, like, do a pie chart of where the progress has come from, it's largely come from massive investment in compute and application of that. I think what's interesting, if I had to, like, if you had gone to my end and said, like, where are we today? I think that we are seeing the flattening off of the value of just adding more compute and more data to language models. Um, I think we've seen truly extraordinary progress over the last five years, but, you know, I, I think of most technologies of S-curves, you know, you sort of get slow progress, then fast progress, then slow, and it strikes me that we're in this moment where sort of incremental value of, of just continuing on this path of scaling is leveling off. And what that means is that the value of ideas is about to go up a lot relative to the value of just scale. And what that means to me is that whenever that ideas become v…

AI assessment note: “value of ideas is about to go up a lot relative to the value of just scale”

Q But that's just really interesting, because everyone is saying, ah, we're seeing the complete commoditization. We even have Chinese players like Yi, I think it's called. Yeah. Which is like, almost equivalent of GPT-IV's performance. Are we not seeing the conversation?

A I think the pure LLM approach absolutely is commoditized. Will be, either is or will be very quickly commoditized. I guess what I'm saying is, I don't think GPT-V is just going to be a bigger language model. I think it's going to involve something different. It's going to involve kind of, uh, you know, like a lot of productizations, maybe search as we already, you know, discussed and things. And so like, but these are not commodity ideas yet. It's not yet, ah, the case that everyone will then just say, ah, or, or that immediately everyone will be able to copy that. And so I think, you know, broadly, you know, I was talking recently to someone who was on the Llama-III team, and they were like, you're probably underring the extent to which Llama-III is literally just Llama-II with more compute and more data. There's not a lot that's different other than, you know, that. I guess what I'm saying is Llama-IV will not just be You know, a bigger Lama three. It will have to be different. It will incorporate new ideas.

AI assessment note: “I think the pure LLM approach absolutely is commoditized.”

Q Do you think the world is ready for an agentic capable model to be unleashed?

A Not one that, where, like, many millions of instances can be run, and, um, they can operate at, like, nearly human level. Now, I don't think that's what GPT-V will be, but, like, I, you know, one of the, one of the theses I'm really interested in right now is, I think we'll get really good agents in the next five years. And I think if you look at parts of the economy where we already have a lot of automation, it's pretty clear that that requires a lot of infrastructure, requires protocols for agents to interact. I think we're going to need that for AI agents. We're going to need, ah, to build great infrastructure to maximize the economic value of these agents.

AI assessment note: “Not one that, where, like, many millions of instances can be run”

Q to founders because it's such a big topic that I do want to kind of really touch on properly, but I want to start on this tweet of yours, which I just thought was fantastic. And you said two predictions. One, someone will build a new open AI, anthropic scale AI company that takes the commoditization of LLMs as one of its core premises. What did you mean by this?

A I was very lucky to start investing in AI about 10 years ago, a bit more than that now, when the first sort of deep learning, uh, wave kicked off. And, you know, what's really interesting is that really, Certainly the last five years, but arguably longer, the story of AI is really a story largely about the deployment of just enormous amounts of compute and enormous amounts of data, and not that many new ideas, candidly. You know, I mean, like, I'm not saying there are none, but, like, broadly, if you had to, like, do a pie chart of where the progress has come from, it's largely come from massive investment in compute and application of that. I think what's interesting, if I had to, like, if you had gone to my end and said, like, where are we today? I think that we are seeing the flattening off of the value of just adding more compute and more data to language models. Um, I think we've seen truly extraordinary progress over the last five years, but, you know, I, I think of most technologies of S-curves, you know, you sort of get slow progress, then fast progress, then slow, and it strikes me that we're in this moment where sort of incremental value of, of just continuing on this path of scaling is leveling off. And what that means is that the value of ideas is about to go up a lot relative to the value of just scale. And what that means to me is that whenever that ideas become v…

AI assessment note: “we are seeing the flattening off of the value of just adding more compute”