The Exchanges

Q What does it mean for the value of OpenAI in Anthropic if model is not the product and it becomes a utility, something you can switch into and switch out of?

A The interface. Everyone thinks we're all building the model layer or the race. We're not actually. Um, I would even argue that building models is a way to stay at the frontier, but you have to own an interface in which valuable AI output tokens are generated, the most valuable tokens. It doesn't have to be the product. This is the single most important thing to, like, You know, unlearn for most founders, and I had to do it too, which is to be successful in AI product layer, whether you're a model builder or not, it's not about building something that gets a billion users. That mentality has to completely shift. There are a few power users who are propelling this token economy right now. If you look at like all these, um, crazy stories of how there's this one engineer who got Amazon spend like half a billion dollars a month because of some stupid way they set up like agent loop inside cloud code. Okay. Maybe that's a mistake, but there are real engineers in meta and in other companies spending like ten million a year per engineer on, on, on, on these, you know, coding tools. There are users in perplexity computer Um, there's one user, I think, who spends upwards of, like, 10,000 dollars a month. Something like that. Crazy. And, and, and not, like, wasting it. They're not wasting money. Their business runs using agent loops that are running inside these harnesses. And they use th…

AI assessment note: “you have to own an interface in which valuable AI output tokens are generated”

Q Do you think Pablassi will be a trillion dollar company?

A Yeah. Anyone can be a trillion dollar company. SK Hynix and Samsung are worth a trillion last, last couple of weeks. Did you know Samsung started off as a grocery store? Did you know that? You didn't know that? Okay. So it's true. They started selling dried fish. Seriously. Um, Hynix was, um, SK, the SK group started off as, um, um, textiles company. So anyone can be worth a trillion dollar company. And like, you just have to work your way towards that. I mean, the exact same logic for you that you laid out for how can a company be worth a hundred billion dollars. Okay. You said you need to make a ten billion dollars in revenue. Isn't that the same for a trillion? Like you need to make a hundred billion dollars in revenue.

AI assessment note: “Yeah. Anyone can be a trillion dollar company.”

Q What does it mean for the value of OpenAI in Anthropic if model is not the product and it becomes a utility, something you can switch into and switch out of?

A The interface. Everyone thinks we're all building the model layer or the race. We're not actually. Um, I would even argue that building models is a way to stay at the frontier, but you have to own an interface in which valuable AI output tokens are generated, the most valuable tokens. It doesn't have to be the product. This is the single most important thing to, like, You know, unlearn for most founders, and I had to do it too, which is to be successful in AI product layer, whether you're a model builder or not, it's not about building something that gets a billion users. That mentality has to completely shift. There are a few power users who are propelling this token economy right now. If you look at like all these, um, crazy stories of how there's this one engineer who got Amazon spend like half a billion dollars a month because of some stupid way they set up like agent loop inside cloud code. Okay. Maybe that's a mistake, but there are real engineers in meta and in other companies spending like ten million a year per engineer on, on, on, on these, you know, coding tools. There are users in perplexity computer Um, there's one user, I think, who spends upwards of, like, 10,000 dollars a month. Something like that. Crazy. And, and, and not, like, wasting it. They're not wasting money. Their business runs using agent loops that are running inside these harnesses. And they use th…

AI assessment note: “you have to own an interface in which valuable AI output tokens are generated”

Q I understand that in terms of the precision around good reasoning. When you think about the trajectory of reasoning quality, I know it's a shit question. Forgive me for it. How do you think about the timeline there? Do you think it goes up, flat, up? Is it a continuous gradual increase? How do you think about the trajectory and slope of reasoning improvement?

A I don't think we know the secret sauce yet. At least according to writers, the, the, the, The report, like, the news media writers, um, they claim, like, OpenAI has the new thing called Qstar, You know, it came out during the whole board saga. Um, and like, that's the sort of thing that they're working on to, like, make these models, like, use their own data to bootstrap and make themselves more intelligent. And then, um, XAI recently hired this guy, uh, Eric Zeligman from, from Stanford, who's written these papers on something called the STAR, um, self-taught automated reasoner. Um, or like self-thought reason, or like, it's basically taking the model itself, make the model explain its own outputs, And then, um, whatever is the right output, you train on that, or it was the wrong output. You take the right output, and then you ask the model to explain why that was right and train on that. So you basically are training on not just the output, but also the explanation that was used to achieve the output. And if you can do that, you're basically training a model that can think and reason and get to an output, see if it's correct. Go back, reason again, and iterate. That is what is lacking in today's models. Today's models are just giving you the output. Tomorrow's models will start with an output, reason, elicit feedback from the world, go back, improve the reasoning, and until t…

AI assessment note: “Maybe it'll be achieved in a year or two. Maybe it'll take three, four years.”

Q Every business has a core monetization engine. They have ancillaries, but there tends to be one which is dominant. When you look at, you know, perplexity in five years time, what is your dominant engine? Is it consumer subscription? Is it advertising? Is it enterprise?

A I would predict it'll be advertising. If we crack it, yes, it'll be advertising. If we don't crack it, if we are not, if we, if we don't, if we haven't grown to that level in user basin, or if we grew and didn't figure out how to advertise really well, I think it'll be the other two. Either way, we can be profitable. I think with advertising, we can be really, really profitable. And then you can ask him, Hey, Arvin, why do you care about profits? Like Sam Altman doesn't care, but he doesn't care because he's not interested in actually just focusing on product as a business. Like he's trying to build AGI. And like he already told publicly in an interview that, you know, even if we spend like, you know, fifty billion dollars on AGI, it doesn't matter. So that's a different company. We shouldn't be seen as an OpenAI competitor at all. We're not an AGI lab. We can say perplexity and ChatGPT are products in a similar space. And, and there's like some competition for mindshare and users, but even that will like be pretty clear, like two years from now, you're not going to keep asking how is perplexity different from chat GPT. Today you are, but two years from now, I don't think so. If that's still the case, one of us is just copying the other.

AI assessment note: “I would predict it'll be advertising. If we crack it, yes, it'll be advertising.”

Q Do you think models are good at reasoning, one, and then I, I think like a breakthrough in reasoning will be one of the biggest kind of, um, breakthrough moments in the next wave. How do you feel about where we are today in terms of quality of reasoning and what is required to break through in the next wave of reasoning quality?

A It really depends on what you call as being good at reasoning. Are, are they better than an eighth grader? I think so. Uh, are they better than like a 12th grader? Are, are, are they better than like 75% of the 12th graders? Most likely. Uh, are they, like, gonna win the IMO or IOI? No, definitely not. So there's, like, a spectrum, right? Of people good at reasoning, even among humans, and I'm sure, like, AI is, like, somewhere, like, in the median right now, uh, of, like, high schoolers. Um, Can it get to like a median college undergrad? Definitely. It seems like we're on the pathway to getting there. Would it be like talking to, um, Faraday or Einstein? Not anytime soon. Uh, I think that's what people are, some people call it as artificial super intelligence. Like, like, um, really the mu plus seven sigma sort of people on the planet. Uh, and like, I think when we achieve that, uh, it'll break all this 20 dollar a month Business models. Like, would you, um, I know, like, you know, businessmen in the past, like, have you watched this movie, uh, Prestige?

AI assessment note: “I'm sure, like, AI is, like, somewhere, like, in the median right now”

Q Where are you still moving too slow internally today?

A I think we can be even more AI-pilled. It's insane. I'm saying this because we are building some of the most interesting AI products and internal adoption of our own products or competitors products can be even higher. And, and, um, this is despite us being extremely, um, Asian build internally and trying to delegate as much to agents. Yeah, that's where that's like a big area for my, my hope is that we can turn this company almost into an AGI. And, and how that doesn't mean no, no humans work here, but there will be an AGI that has all the context it needs to run different divisions of the company. In a semi-autonomous way with some scaffolding provided by humans here and there. And, and that's not, that's going to feel, that's not going to feel scary at all. We'll normalize that, that feeling very fast. It's just going to feel like, and you know, uh, the X engineers running certain aspects of the company.

AI assessment note: “internal adoption of our own products or competitors products can be even higher”

Q How important is it that we have our own TSMC in the US?

A So TSMC is actually, there is a fab of TSMC in Arizona. Like not a lot of people talk about this, but TSMC is investing like hundred, fifty billion dollars into that, into, into building American fabs. And, um, they've already invested forty billion dollars or something like that. Sixty billion last time I checked. So there is a TSMC in Arizona that's coming up. There's also, um, Intel. And that's why, you know, American government owns 10% of Intel. Nvidia and SoftBank own five percent each. So there is a lot of investment going into an American fab as well as TSMC is investing into its American fabs. Elon's building Terra fab. Like, I think people have woken up to the importance of building fabs. But, um, this is also why China is particularly very, very competent because

AI assessment note: “I think people have woken up to the importance of building fabs.”

Q Are you not worried about the wealth inequality? Yeah, Aaron, if we were being blunt, we both are very lucky now to live in kind of nice worlds and rarefied heirs. Are you not worried by just how much money a very small number of people have, and how fucking hard it is for everyone else, and that gap is getting bigger?

A I think the way to like ensure that that's not, that doesn't remain the case is to distribute the benefits more widely. You got, you got to let any, by the way, the people who are using our tools, like I had an Uber driver, I'm not even like making this thing up. So, um, as honest as I can get, an Uber driver in San Francisco, um, once told me that he watched one of my, uh, YouTube interviews Where I explain how you can build a product or a web app with an AI from scratch. I went on to do it and, um, um, use AIs to add like billing and all that. And that makes more passive income for him than, um, driving Ubers. And so he actually reduced the amount of time he's driving Uber because he, he loves live coding new apps. And, uh, that, that already tells you that For the person with agency and a positive outlook for the future, anything is possible. And so if you keep communicating all the negative things you can about AI and wealth inequality all the time, and that's the only thing news and press writes about, I think it'll perpetuate and people will only think the bad things. And so it's, it's, it's very essential that if you think you're already doing well, it's very essential that you talk about what are all the things that can go well and give hopes to people who were once upon a time like you, like you, I mean, you were, you didn't, you, you started this, um, podcasting circu…

AI assessment note: “the way to like ensure that that doesn't remain the case is to distribute the benefits”

Q is a, is a big beast to get your head around, um, and it's a challenge. You know, you, you said there about the scale of OpenAI sales team. I know, I've got many friends in it. It's a big thing. How do you think about getting your head around the GTM building exercise of an enterprise division? Do people buy perplexity enterprise and open AI enterprise or either or?

A My sense is that like AI is still so early today that nobody's locked in and loyal to any, any particular, uh, enterprise tool in AI and none of them even have a lock in effect. To, like, make your data live on, like, one single, uh, tool. Like, it's not, I'm not even talking about things like, why is it hard to migrate from Snowflake to Databricks? Because the SQL format itself is so different, and once you wrote all the SQL queries in one format, it's so hard to change. Like, like, it's not even things like that in AI, like, your custom prompts that you wrote for Chatshipt can be taken over easily to perplexity. It's very easy. So, um, I, I think enterprises are still willing to tinker and experiment and try different tools. And that said, if there is no differentiation, they will win. In the beginning, the one with the bigger brand and bigger team has an advantage. But is it game over? No. It's just game begins today. I think, like, this is exactly the whole wrapper thing, and I, um, if the value you add is, like, uh, very little on top of the model, or if the model is the one that's adding most of the value and all the stuff you built around it don't matter, yes. But if you build enough value around the model that is very difficult to do without coordinating a bunch of Other hard to achieve engineering feeds that are not just LLM spaced or like have a lot of human element i…

AI assessment note: “enterprises are still willing to tinker and experiment and try different tools.”

Q What do you think is the best question you are never asked? You've done interviews before.

A I think someone asked me, um, like, why are you doing this sort of thing? And There's a sort of question where you don't actually know yourself. I think a lot of people give, uh, these, um, made up answers. Like, oh, I had an existential crisis. I needed to save humanity from extinction. Uh, like, you know, I need to preserve the light of consciousness. And so I thought about what are the most important, like, these are sort of things that I've seen entrepreneurs say, but, uh, reality is like, like, you just sort of look up to some people, you want to be like them, and you try to Carve your career path according to what they have done. But then you end up like figuring out there are things that you really like and you shape it to the style you want. And at least that's how it's been for me. I have been a big fan of Larry Page and I always wanted to do some things that, of that, of that scale of ambition. But that was not the reason we did search engine though. Like we, we started with something else completely. So that's a question that I actually don't have a clear answer to, but I really liked the question because it's a question worth asking yourself constantly. Like, why are you even working on this? Like, uh, Steve Jobs has this thing, right? Like if you, uh, felt like if you, if you, uh, internalize death, if you normalize death and, and every day morning you stood in fro…

AI assessment note: “I think someone asked me, um, like, why are you doing this sort of thing?”

Q Where are you still moving too slow internally today?

A I think we can be even more AI-pilled. It's insane. I'm saying this because we are building some of the most interesting AI products and internal adoption of our own products or competitors products can be even higher. And, and, um, this is despite us being extremely, um, Asian build internally and trying to delegate as much to agents. Yeah, that's where that's like a big area for my, my hope is that we can turn this company almost into an AGI. And, and how that doesn't mean no, no humans work here, but there will be an AGI that has all the context it needs to run different divisions of the company. In a semi-autonomous way with some scaffolding provided by humans here and there. And, and that's not, that's going to feel, that's not going to feel scary at all. We'll normalize that, that feeling very fast. It's just going to feel like, and you know, uh, the X engineers running certain aspects of the company.

AI assessment note: “internal adoption of our own products or competitors products can be even higher”

Q Given the capabilities of China that we just mentioned that really articulately, I know it's a ridiculous question, but, um, if I were to say to you, your job is to make sure America stays competitive, what would you do to ensure that you retained competitiveness in an increasingly strong China?

A I think, I think take physical infrastructure a lot more seriously and continue funding it. Um, and not like how all these, you know, I, I wouldn't say meaningless. It's more like not propagate fake news around data centers, um, about how data centers are polluting and contaminating water, or like they're sucking up all the water. Um, and, and, and actually be fact driven. And so, you know, I hope our product helps there. Like you, you can, you can go to perplexity and ask any question and get fact checked on your assumptions. But yeah, like, it's very important that we educate the public, um, about what's actually going on in, in a language they easily understand and not fear monger. Okay. Not be like, oh, like all their jobs are going to go away. Like there's that, like there's going to be lots of amazing companies that are going to get built with far fewer people getting multi-billion dollar, multi-hundred million dollar valuations with like 20, 30 people and propelling like trillions of dollars of new GDP. Like let's talk about how to enable that. Let's talk about how to build that and create a more positive Future together, right? Uh, instead of, oh, like, 90% of the jobs are gonna be gone. Like, you're all gonna get screwed over by our models, and like, and, and it's, it's our, it's our moral duty to tell you all this. Like, blah, blah, blah. Like, that doesn't make any s…

AI assessment note: “take physical infrastructure a lot more seriously and continue funding it”

Q What do you think is the best question you are never asked? You've done interviews before.

A I think someone asked me, um, like, why are you doing this sort of thing? And There's a sort of question where you don't actually know yourself. I think a lot of people give, uh, these, um, made up answers. Like, oh, I had an existential crisis. I needed to save humanity from extinction. Uh, like, you know, I need to preserve the light of consciousness. And so I thought about what are the most important, like, these are sort of things that I've seen entrepreneurs say, but, uh, reality is like, like, you just sort of look up to some people, you want to be like them, and you try to Carve your career path according to what they have done. But then you end up like figuring out there are things that you really like and you shape it to the style you want. And at least that's how it's been for me. I have been a big fan of Larry Page and I always wanted to do some things that, of that, of that scale of ambition. But that was not the reason we did search engine though. Like we, we started with something else completely. So that's a question that I actually don't have a clear answer to, but I really liked the question because it's a question worth asking yourself constantly. Like, why are you even working on this? Like, uh, Steve Jobs has this thing, right? Like if you, uh, felt like if you, if you, uh, internalize death, if you normalize death and, and every day morning you stood in fro…

AI assessment note: “I think someone asked me, um, like, why are you doing this sort of thing?”