Dylan Patel: GPT4.5's Flop, Grok 4, Meta's Poaching Spree, Apple's Failure, and Super Intelligence

Super intelligence reaching it first. Who are you picking and why? OpenAI. He's the guy the chip industry reads before making a move. Meet Dylan Patel. He's a quick thinker with a depth and breadth of knowledge that is unrivaled in AI.

You know, for one, scale AI is like it's kind of cooked. And today, Dylan's answering the tough questions. What went wrong with GPT 4.5? In general, it's not that useful and it's too slow. You zoom back, right? It's like if you believe super intelligence is the only thing that matters, then you need to chase it. Otherwise, you're a loser. It's not the money, it's more the power.

What do you think is going on at Apple? They hate Nvidia. Maybe for reasonable reasons. Models are just like pansies about like giving me data. You're using 03 day-to-day even though there's, you know, it takes so much time to actually get your response back. The model I go to the most is either 50% of white-collar jobs could disappear. Generally, people work less than ever before. The average amount of hours worked 50 years ago was way higher. And then eventually, like there just won't be humans in the loop, right? And eventually, like you believe that it's an art form to some extent. What is worth researching and what's not? GPUs that ended up breaking. It was called bumpgate. It was a very interesting thing. Um, you don't or you do. No, I don't. Oh, so okay. So this is a very fun story, right?

All right, Dylan, thank you so much for joining me today. Really excited to talk to you. I've seen you do a number of talks. I've seen you do a number of interviews. We're going to talk about a whole bunch of things. So first thing I want to talk about is Meta. Let's start with Llama 4. I know it's been a little while in the AI world since that kind of released, but there was a ton of anticipation. It was good, not great. It wasn't kind of world-changing at the moment. Um, and then they delayed Behemoth. What do you think is going on there?

Yeah. So, I mean, it's funny. There's like three different models, and they're all quite different. So, Behemoth got delayed. Uh, I actually think they might not ever release it. I'm not sure. There's a lot of problems with it; the way they trained it, some of the decisions they made don't pan out. And then there's Maverick and Scout, right? Um, and so one of those models is actually decent. It's pretty good. It wasn't the best on release, but it was it was comparable to the best Chinese model on release, but then, you know, Alibaba came out with a new model. Deepseek came out with a new model, so I was like, "Okay, it's worse." Uh, the other one was objectively just bad. I know for a fact they they trained it as a response to DeepSeek, trying to use more of the elements of DeepSeek's architecture, but they didn't they didn't do it properly; it was just a rush job and it really messed up um because they went really hard on the uh on the sparsity on the E. But funnily enough, if like you actually like look at the model, it often times won't even route tokens to the certain experts. So it was like a waste of like training, basically like you know like within in between every layer the router can route to whatever expert it wants to, and it learns which expert to route to um and and each expert learns its like own independent things and it's like really not something observable by people, but what you can see is tokens when which experts do they route to like or you know when when they when they go through the model, and it's just like some of them just didn't get routed to. So it's like you have a bunch of empty experts that are not doing stuff. So there's clearly something wrong with training.

Is it like an expertise thing internally? I mean, they have to have some of the best people in the world, and we're going to get to some of their hiring efforts as of late, but like what what like why why haven't they been able to really do it? I think it's a combination and confluence of things, right? Like yes, they have tons of talent. They have tons of compute, but the organization of people is always like the most challenging thing. Which ideas are actually the best? Who's the technical leader choosing the best ideas? Right? It's like if you have a bunch of great researchers that's awesome, but then if you put like product managers on top of them and then there's no technical lead who's like evaluating what to choose then you have a lot of problems, right? OpenAI, right? Yeah, Sam is a great leader, and he gets all the resources, but the technical leader is Greg Brockman, right? And Greg Brockman is choosing a lot of stuff, and there's a lot of other folks, right? like Mark Chen and others who are like the technical leaders who are like really deciding the like you know technically which route do we go down because a researcher is going to have their research, they're going to think their research is the best. Who's evaluating everyone's research and then deciding that idea is great, let's use that one. That one sucks. Let's not use that one. It's just really difficult. So, when you end up with researchers not having a leader who uh is technical and can choose uh and and really, you know, choose the right things, you end up with great, we did have all the right ideas. Um, but part of AI research is that you have all the wrong ideas too, and you learn from them and you have the right ideas and you choose them, right? And now what if what happens if your choosing of them is really bad and actually you choose some wrong ideas and then you go up the, you know, the branch of sort of research, right? Like you've chosen this bad idea. This is something we're going to do. Let's go further down. And then now you're like, oh, branching off of this bad idea, there's a lot more research because, you know, it's like, well, we're not going to go back and undo the decision we made, right? Right? So everyone's like, "Oh, we made that decision. Okay, let's see what's researchable from here." And so you end up with like this like potentiality of like great researchers wasting their time on bad paths, right? And and there's sort of this like thing that researchers talk about which is taste, right? Which is very funny, right? You think like these are like these are like nerds who won like the International Math Olympiad and like that's their like, you know, claim to fame when they were like a teenager and then they got a job at OpenAI or whatever at 19 um or a Meta or whatever. But there's actually a lot of taste involved, right? It's uh it's it's it's an art form to some extent. What is what is worth researching and what's not, and it's an art form of choosing what's the best because you're making all these ideas down here on the scale here and then all of a sudden you're like yeah let's now you know great those experiments were all done with like 100 GPUs. Awesome. Now let's make a run with 100,000 GPUs with that idea. It's like well things don't just translate perfectly. So there's a lot of taste and intuition here. It's not that they don't have good researchers. It's that like who's choosing the taste, right? Is is difficult, right? Like, you know, it's like you don't care about movie critic reviews, you care about Rotten Tomatoes, you know, audience score perhaps, right? And it's like it's like who's the critic that you're listening to, though, right? And that's that's it's it's challenging to you know even if you have great people to actually have good stuff come out because of organizational issues because the right people aren't at the right spot and decision makers and maybe the wrong person gets to like be political and have their idea and research path put into the model when it's not necessarily a good idea.

Yeah. And okay, well let's continue down the path of who is making decisions. Obviously Zuck last week there was a lot of uh news about him giving a hundred million offers. I mean, Sam Altman literally said it. They acquired Scale AI seemingly for Alexander Wang and his team. He's in founder mode. What does the Scale AI acquisition actually give Meta first? Let's start there.

Yeah. So I think um you know, for one, scale AI is like it's kind of cooked right now as a company as a company because everybody's cancelling their Google Google's backing out. I think they're going to spend on the order I've heard like $250 million this year with them, and they're backing out. Obviously, they've spent a lot of money. There's stuff they can't back out of. But it's like that's going to go down a lot, right? OpenAI allegedly like cut the external Slack connection, right? So there's no like Slack between Scale and OpenAI anymore. So obviously that ultimate breakup between companies. Yeah. So like obviously these companies like I don't want Meta to know what I'm doing, right, with my data because that's one of the unique aspects of models is like what do you what do you want with your custom data? Um so clearly Scale is not you know Meta didn't buy Scale for Scale, right? um they bought Scale for the purposes of having Alex and his few best uh colleagues the there's a few other folks at at at Scale who are really awesome as well, and and they bought them they bought them to bring them over, right? um now the question is sort of like is the data that Scale has good is is all knowing all the paths of sort of data labeling that all these other companies were doing good. Sure. Um, but more importantly, it's like we want to get, you know, someone to help us lead this super intelligence effort, right? And and Alex is a uh, you know, he's he's same age as me. He's 28 or 29ish yet. I think he's he's pretty he's around that age. Stupendously successful in every way, shape, or form, right? People can hate on him if they want, but he's obviously very successful, especially when he convinces Mark Zuckerberg, who's not an irrational person, he's very smart, to buy his company, right? like you know it's like and there his company was doing you know nearly a billion of revenue and is like let's chase super intelligence right which is very different right if you go look at Zuckerberg's interviews even a handful of months ago he wasn't chasing super intelligence right he was chasing like AI is good and great but like AGI is not a thing that is going to happen soon um so sort of this is a big shift in strategy um in that he's like basically saying yeah yeah like super intelligence is all that matters we're on the path there I believe now what can I do to get catch up cuz I'm behind.

It seems like the narrative throughout all of these major companies is now super intelligence even when it was AGI just a month ago. What, why the transition by the way? The word AGI has no meaning anymore. Amorph is. Yeah. Right. It's like it's like you can like look at an anthropic researcher in the face and be like what does AGI mean? And they literally think it just means an automated software developer. And it's like that's not artificial general intelligence, but that's what they think, right? uh and like a lot of researchers across the ecosystem. Ilia saw this, you know, Ilia saw everything first, obviously, Ilia Suster. Um, and he started his company, Safe Super Intelligence, right? SSI, and I think that started the rebranding. And now, like, you know, many months later, it's like almost a year now. A year later, um, I think it's like nine months to a year later, you know, everyone's like, "Oh, super intelligence is a thing." So, another another research direction, right, that Ilia got first, right? Whether it was like whether it was like pre-training scaling or like uh, you know, the original like vision networks, right? pre-training, scaling, reasoning, right? All these things that he sort of had the idea at least if not first among the first and and uh worked on it a lot. Um you know, sort of Ilia's got this one too, which is the rebranding. So maybe he's got marketing, too.

Yeah. Well, Zuck, I at least rumored tried to acquire SSI and was rebuffed by Ilia, right? And then um I wanted to also ask you about Daniel Gross and Nat Friedman. I think it's rumors maybe confirmed at this point, but seems like Zuck is trying to hire them as well. What did those two folks give Zuck? Zuck tried to buy SSI. He also tried to buy Thinking Machines. Like these are rumors. He also tried to buy Perplexity. These are all in some of the media, right? He tried to buy all these companies, but specifically like some of the rumors that have been floated around is basically that like Mark tried to buy SSI. Ilia obviously said no because he is like committed to super intelligence and straightshotting it, right? Not like worried on products and he's probably not even that money focused, right? He's mostly focused on like um building it, right? Um a true believer in all all respects and regards, right? Um so obviously he probably was like no. Um I don't know what the makeup of equity is there, but Ilia's Ilia's probably got strong enough votership and ownership to be like no. Um, and if the rumors are true about Daniel Gross, then like Daniel Gross probably was the one wanting the acquisition, right? He's like, "Yeah, this is awesome." Yeah. Another founder. Um and he comes from you know not an AI research background although he's is he's technical to some uh to a degree but like you know it's like he you know he had his venture fund with Nat and he then he founded SSI with Ilia and he probably wanted the acquisition and and and then it's like well I was pushing for an acquisition and it didn't happen and you know I'm just guessing like you know he he's he's probably if he if he's going at all I don't actually know if he's going at all. Um it would make sense that like that's a chasm and split and he's going um and I think I think generally when you like look at really uh a lot of people who are very successful um it's not the money right it's is the money always but it's it's more the power right um and if you ask like you know anyone going to Meta a lot of them will obviously be going for money but a lot of them are going because now they have control over uh the AI path for, you know, a trillion dollar plus company and they're right there talking to Zuck u and they can convince one person who has full voting rights over the entire company, right? It's like there's a lot of power there, right? And they can implement across billions of users, right? Whatever AI technology they want using the engine of Facebook, whether it be infrastructure or researchers or product to like push whatever AI product you want, right? And that like that would like make a lot of sense to me for like an Alex Wang or a Mark Zucker or for a for a Nat Friedman or a Daniel Gross who are um they are much more product people right like Nat doing GitHub co-pilot he's a product person right um he's not an AI researcher although he knows a lot about AI researcher he's a he's a product person right and and same applies to sort of like Alex like obviously he's he's very well versed with the research but his super skill set is is product and people and like convincing people and organization probably not as much the research. That's sort of the the angle there is like they've got they've got like all this power to do a lot at Meta.

Sam Altman also mentioned that Meta has been giving $100 million bonus offers to their top researchers. Apparently, none of the top researchers have left. I I want to ask is is that a successful strategy just to like throw money at the problem, get the best people in? It feels like maybe the the cultural element would be lacking there where you know give as much as you want to OpenAI and Sam Altman but they like there are a lot of true believers there in what they're doing. Is that enough to just throw money and get the best researchers where that culture is going to be built? You zoom back, right? It's like if you believe super intelligence is the only thing that matters, then you need to chase it. Otherwise, you're a loser, right? And Mark Zuckerberg certainly doesn't want to be a loser and he thinks he can build it. He can build super intelligence too, right? So then the question is like, okay, well, what do you do? Well, then you go and try and acquire the best teams out there, right? Thinking machines, right? All these ex-OpenAI people, but also there's some other folks from, you know, Character AI, GDM, Meta, etc., right? All these great researchers and and infra people. And same with SSI, right? It's Ilia and the people he's recruited. Um, you know, trying to recruit people from these companies, uh, or try and buy these companies. That didn't work out. Um, so now you go with like Alex who's like tremendously like connected and can help you build the team and now you got to go get the team right now. What's the difference between like acquiring SSI where there's like way less than 100 employees, right? You know, I think I think there's less than even 50 employees at SSI and for for $30 billion and and like okay, well, you just paid hundreds of millions of dollars per researcher and you know 10 billion plus for Ilia, right? Like that's sort of what you just did. And it's like well then you're sort of doing the same thing, right? As far as Sam is saying that no top researchers have gone, I don't believe that's accurate. Um I think I think initially the top researchers definitely did say no. The best researchers, the best people. Um, and and you said $100 million dollars. I've I've heard a number for someone over a billion actually, uh, for one person at OpenAI. Um, but anyways, you know, that's it's a ridiculous amount of money, but it's like, well, it's the same thing as buying one of these companies, right? Thinking machines and SSI don't have a product. You're buying them for the people. If you know, super intelligence is the end all be all, $100 million, even a billion dollars is really a drop in the bucket compared to one Meta market cap currently. And also the total addressable market of artificial intelligence.

I want to talk about Microsoft and OpenAI's relationship a little bit. We're well past the honeymoon phase, it seems. It definitely seems to be kind of the choppy waters of their relationship now. It's clear that open this is now a therapy show. Yeah, absolutely. Tell me about your feelings, Sam and Satya. Well, this is therapy, right? These are two people, and they have a relationship, and it does seem to be folding a bit. OpenAI's ambitions seem to have no bounds. Is Microsoft thinking right now? They they want to restructure the deal. OpenAI does Microsoft really has no reason to. But like what do you think is going on at like what do you think about the dynamics of this relationship going forward? Like OpenAI would not be where they are without Microsoft, and and Microsoft signed a deal where they get tremendous power. It's a weird ass deal because like OpenAI wanted to be a nonprofit and they cared about AGI, but at the same time they um had to give up a lot to get the money. Uh, but at the same time, Microsoft didn't want to run into antitrust stuff, so they like structured this deal really weird, right? Which is like there's like revenue shares and there's like profit guarantees and there's like all these different things, but nowhere is it like, oh yeah, you own x% of the company, right? I think it's like they get like I don't it's off the top of my head but I think it's like 20% revenue share 49 or 51% like um profit share up until some cap and then there's like Microsoft has the IP rights of all of OpenAI IP until AGI, right? AGI and it's like all of these things are just like nebulous as hell, right? It's like I think I think the profit cap might be like 10x again I'm like going off the top of my head it's been a while since I looked at it but it's like if if Microsoft gave roughly $10 billion and OpenAI has it's it's a profit cap of 10x. It's like well like what incentive does Microsoft have to renegotiate now if they get a hundred billion dollars of profit from OpenAI and until then OpenAI has to give them all their profit or or half of their profit right and they get this 20% rev share and they have access to all of OpenAI's IP until AGI but like what is the definition of AGI? Like theoretically, Microsoft OpenAI's board gets to decide when OpenAI hits AGI, but

Then, if that happens, Microsoft will just shoot, sue the out of them. Um, and Microsoft has more lawyers than God. Um, so it's like, it's like this just like crazy ass like deal. I think there's like a few like really worrisome things in there for OpenAI. It's like one of the main things they got ex-removed already because Microsoft was really scared, I think, about antitrust aspects of this, which was that OpenAI had to exclusively use Microsoft for compute. Um, that that they backed off of this last year, uh, and it got announced with the Stargate deal this year, right? Which was that um OpenAI is going to go to Oracle and uh SoftBank and uh Crusoe and uh the Middle East to build their Stargate clusters, right? Their next generation data centers. They're still getting a bunch from Microsoft, of course, but a bunch from uh Stargate. And so, you know, or from Oracle primarily, but the others as well.

Whereas before it was that OpenAI could not do that without going directly to Microsoft, right? OpenAI tried to go to CoreWeave initially, but then Microsoft sort of inserted themselves in the relationship like, "No, you're exclusively using us." So a lot of GPUs got rented from CoreWeave to Microsoft to OpenAI, um, but then this like exclusivity ended, and now like CoreWeave has big deals signed with OpenAI and Oracle has big deals signed with OpenAI. What did Microsoft get in that exchange where they're going to give up the exclusivity? Did they get anything, or was it reported that they got anything in exchange for that? Usually, it's not just like, "Okay, cool, we'll give that up." What's been reported is just that they gave up the exclusivity, and in return all they have is a first-right-of-refusal. Anytime OpenAI goes and tries to get a contract for compute, Microsoft can provide that same compute at the same price in the same time frame because it reduces risk from antitrust. Yeah. Like antitrust is one of the biggest considerations there, but there's other considerations, of course, but like antitrust being one of the big ones, cuz like being the exclusive compute provider to OpenAI is a little, little uh iffy.

Um, and and from OpenAI's perspective, they were just really annoyed that Microsoft was way slower than they needed them to be, right? They just couldn't get all the compute they needed. They couldn't get all the data center capacity, etc. Um, CoreWeave and Oracle are moving much faster. Um, but even they are not as fast, and so OpenAI is turning to other folks as well, right? There's that butting of heads there, but nowadays like the real challenging thing here is like Microsoft has the monorepo; it has the OpenAI IP; they have rights to it all; they can do whatever they want with it. Um, now, whether it's like Microsoft playing nice and not doing stuff with it or being somewhat incompetent and not being able to leverage it and mostly just like looking through it, um, whatever the reason is, Microsoft, you know, despite having access hasn't done a ton, but the possibilities are endless, right? Like, of what could be done.

Then the other thing is like if you're truly like AGI or now superintelligence pilled, you have all the IP up until SS super intelligence is, is let's just say achieved, but that would imply that like the day before super intelligence is achieved you have all of the IP, and then it gets cut off, but you have all the IP up until right there. So it's like one day of work, maybe it's hard, and it takes 10 days of work instead of one. Or maybe you achieve super intelligence, but it takes some time to get to the deliberations and agreement that you've achieved super intelligence/all the evidence that you've achieved super intelligence, but like you've claimed it at this date, but like the model that is super intelligent is here; like you made it here. OpenAI, Microsoft has access to it, right? So sort of that's the real big risk or sort of to the super AGI pilled uh folks. The profit share and all this is like very clean and difficult, and most people don't care that much during when they're investing in OpenAI. It is challenging to get every investor in the world to be like, "Yeah, your crazy ass structure, nonprofit, for-profit, all this sort of stuff." Okay, that's fine. Oh, Microsoft has rights to all of your profit for a long time and all your IP. So, theoretically, you could be worthless if they decide to just like take some of your best researchers and implement everything themselves. Oh, wow. Right? Like these sorts of things scare investors, and Sam said it himself, OpenAI is going to be the most capital-intensive startup in the history of humanity. Yeah. Right.

Um, the valuation is going to keep soaring because of what they're building. OpenAI has no plans to produce profit anytime soon. They've been around for so long, and they're doing like $10 billion of revenue, and they're still not going to do profit for another five years. And and by then their projections of pro of revenue are like well north of like, like there are hundreds of billions if not trillion dollars, right, of revenue that they expect themselves to have before they ever turn a profit. And so that whole way through they're going to be losing money, and they need to keep raising money, and they need to be able to convince everyone who's an investor in the world. And like these things are dirty, right? Like they're not they're not clean and easy to understand. Okay. So you talked a little bit about compute capacity and specifically with Azure being able to go to CoreWeave and and elsewhere. I want to talk specifically about GPT-4.5. It was deprecated, I believe, last week. Um, this was a a massive model, much—was it really? It wasn't it? Oh, I don't know. I thought it was still available in chat. That's I was just curious. Oh, uh, maybe they just announced the deprecation, but it was imminent. No, it's still there, but yeah, they announced it. Okay. Yeah. No, no, they they've talked about like no use. There's very little usage of it, so that makes sense.

Was the model too big? Was it too costly to run? Was like what went wrong with GPT-4.5? Orion, as it was like internally called, u what they hoped would be GPT-5. They made that bet like in early '24, right? They started training it in early '24. It was a bet on full scale, right? Uh, full-scale pre-training, right? We're just going to take all the data. We're going to make this ridiculously big model, and we're going to train it. It is much smarter than 4.0 and 4.1, to be completely clear. I've I've said it's the first model to make me laugh like because it's actually funny. Um, but in general, it's not that useful, and it's too slow and it's too expensive. Um, versus other models, right? Like 3.0 is just better. They went pure on the pre-training scaling; data doesn't scale right, so they weren't able to get a ton of data. So without data scaling so fast, right, they have like this model that's really, really big, trained on all this compute, but you have this issue called overparameterization. Generally, in machine learning, uh, if you build a neural network and you feed it some data, it will tend to memorize first, and then it will generalize, right? Um, i.e., It'll know that if I ever say, "The quick brown fox jumped over," it would just know the next token is always lazy. Right? It isn't until you've trained it on a lot more data that it learns what quick brown fox even means, what lazy dog is, right? It doesn't it doesn't actually build a world model. It generalizes. Um, and to some extent, GPT-4.5 Orion was so large that it was and so overparameterized that it did—it memorized a lot. Like, actually, when it initially started training, I know people at OpenAI were so excited that they were like, "Oh my god, it's already crushing the benchmarks." And we're barely into training because some of the checkpoints were just so good, right? Initially, but that's because it just memorized so much. Um, but then it stopped improving. It like was just memorized for a long time, and it didn't generalize. It finally did generalize, right? Because it was such a big complicated run, they actually had a bug in it for like months, right? Uh, during the training, right? Training is like like usually a handful of months or less, right? And it's usually less. Um, and they had a bug in the training code for a couple of months that was like a very tiny bug that like was messing up the training. It's funny like when they finally found it, it was like it was like a bug within PyTorch that like OpenAI had like had found and fixed, and they submitted the patch; there's like 20 people at OpenAI who reacted to the bug refix reaction with like emojis, right, on GitHub.

Another thing is like they had to restart training, you know, from checkpoints a lot; it's so big, so complicated; so many things can go wrong, right? And so from an infrastructure perspective, just corellating that many resources and putting them together and having it train and having it train stably was really, really difficult, but from another flip side, it's just like even if the infrastructure and code and everything like that was pristine, you still have this problem of data; you know, you're you're you sort of like—everyone points to the Chinchilla paper uh from '22, I I think, uh 2022; Google released a paper called Chinchilla, DeepMind, u and what it basically said is like for a model, what's the optimal ratio of parameters to tokens, and this only applied to dense models with the exact architecture of the Chinchilla model, um, but it was like, oh, if I have like x FLOPS, I should have this many parameters, this many tokens, right? It's a scaling law, right? Um, obviously, as you make it bigger and you apply more FLOPS, the model gets better, but how much data should I add? How much more parameters should I add? Now, obviously, so so over time, you know, people's architectures change; the exact observations of Chinchilla aren't accurate, right? Which is that like roughly it's 20 tokens per parameter that you want of data that you're training versus parameters in the model, roughly. Um, there's actually a curve and everything; it's more complicated than that, but like that observation is not like identical, but what what it is is that like as you add compute, you want to add more data and parameters at a certain ratio or along a certain curve of like, you know, you know there's there's a formula basically, and in an ideal world, and they didn't go there, right? They had to go to way more parameters versus tokens, but this was all early '24 when they started training, um, you know, all these trials and tribulations; they finally get there, and I don't remember when they released 4.5; it was last year, right? Yeah, yeah, but they finally released the model, um, you know, many months after they start training, after they finish training pre-training, and then they try and do RL and all this stuff, but in the meantime, different teams at OpenAI figure out something magical, which is the reasoning stuff, the Strawberry. So it was it was—while they've already invested all this, while they're in the process of training this massive model, they were they realize, "Okay, for a much lower cost, we can get so much more efficiency, so much higher quality out of a model because of reasoning." Yeah. And if you really like try and boil down reasoning to first principles, you're giving the model a lot more data. Where are you getting this data from? Is you're generating it. And how are you generating the data? Well, you're creating these verifiable domains where the model generates data, and you throw away all the data where it doesn't get to the right answer, right? Where it doesn't verify that that math problem or that code or that unit test was good. So in a sense, it's kind of like, you know, like, you know, looking backwards; obviously, I didn't have the intuition then, but like looking backwards, the intuition makes a lot of sense that like, well, 4.5 failed because it didn't have enough data; also, it was just very complicated and difficult on a scaling perspective, infrastructure-wise, um, and there were tons of problems and challenges there, uh, but also it just didn't have enough data, and now like this breakthrough that happened from a different team is generating more data, and that data is good, right? Like a lot of the synthetic data stuff is like bad data, um, but like the the magic of Strawberry, you know, of of reasoning is that the data is good. The the data that you're generating. So it's it's it's really like from a first-principles basis, makes a lot of sense that data is the wall. U you know, just adding more parameters doesn't do anything.

I want to talk about Apple for a second. Um, I'm sure you have some thoughts on that. Apple is clearly behind. We're not getting much in the way of public models, leaks, anything about knowing what they're doing. What do you think is going on at Apple? Do you think they just made a misstep? They kind of were late to the game. Why aren't they acquiring companies? Like, what what is happening internally if you had to guess? Yeah. So I think Apple is like a very, very conservative company. They've acquired companies in the past, but they've never done really big acquisitions. Beats was the biggest one, as a headphone company, right? But generally, their their acquisitions have been really small, and they do buy a lot of companies. They just buy really, really small companies, um, early. They identify it. Maybe it's a failing startup, or it's, you know, whatever it is. They buy they buy these startups that haven't achieved product-market fit and aren't like super sexy. Like, as far as like Apple, they've always had problems attracting, in terms of AI researchers, AI researchers like to blab. They like to post and publish their research. Um, and Apple's always been a secretive company. They actually changed their policies to where their AI researchers are allowed to publish. Uh, but at the end of the day, they're still a secretive company. Um, they're still like an old, antiquated company. It's like like Meta only was able to like hire a bunch of researchers and talent because they had a bunch of ML talent already, right? They've always been a leader in AI. They had this PyTorch team as well. And then they're they committed to open-sourcing a lot for a while now. They've been open-sourcing. Yeah.

Besides that, like who's been able to acquire AI talent? The DeepMind to OpenAI shift, you know, OpenAI being like the rival to DeepMind and like that whole thing and like all a lot of great researchers coming together to form it, and then the Anthropic splinter group and then like Thinking Machines splinter group from OpenAI and SSI's thinking uh splinter group from OpenAI, right? It's like what companies have actually been able to acquire talent that didn't already have AI talent? Like Google DeepMind is just like the biggest name in the game, and they've always had the highest inflows of AI researchers and PhDs, um, and then there's like OpenAI and Anthropic who are sort of and and Thinking Machines and SSI, right? It's all OpenAI; it's hard to get talent to come to you now. Anthropic has such a strong culture that they're able to get people. OpenAI is being the leader. Uh, Meta, you know, I sort of talked through it; it's like how is Apple going to attract these best researchers? They're not, right? They're going to get, you know, not the best researchers. Um, and so it's it's really challenging for them to be competitive, right? And then there's the whole like they have a stigma against—they hate Nvidia. Um, and like maybe for reasonable reasons. Um, you know, Nvidia threatened to sue them over some patents at one point. Um, Nvidia um sold them GPUs that ended up breaking. It was called Bumpgate. It was a very interesting thing. I don't remember that. Um, you don't or you do? No, I don't. Oh. So, okay. So, this is a very fun story, right? One generation of Nvidia's GPUs. I'm I'm going to butcher the exact reason because it's been a while since I ago was this. Uh, this is like this is probably like 2015, if not earlier. Um, there's a generation of Nvidia GPUs for laptops, right? Um, and and chips have solder balls on the bottom, right, that connect their IO pins to the motherboard and you know to the CPU, power, etc. Somewhere along the chain, supply chain, all the companies—Dell, HPE, Apple, Lenovo—they blamed Nvidia as far as I understand, but vice versa, Nvidia said it wasn't their fault. I'm not going to prescribe blame, but like the solder balls would not—like they were like not good enough, right? And so when the temperature is swung up and down, uh, coefficient of thermal expansion, right, different materials expand and shrink at different rates, um, the chip versus the solder balls versus the PCB would expand and shrink at different rates. And what ended up happening is um because of that different rate of expansion, the solder balls connecting the chip and the board would crack. They would, and it was called Bumpgate. U and now the connection is severed, right? The connection between the chip and the board. And so it's called Bumpgate. And um I think Apple wanted compensation from Nvidia. I think Nvidia was like, "No." Uh, there's this whole thing; Apple really hates Nvidia because of that and because of this like threatening uh when Nvidia was trying to get into mobile chips because they tried to get into mobile chips for a time period and they failed. Uh, but at one point they tried to sue everyone uh over over GPU patents in mobile. Um, and so between those two things, Apple really doesn't like Nvidia, and so Apple doesn't really buy much Nvidia hardware. They don't really need to anymore. Well, they don't need to in the laptops, of course, but like even in data centers, right? It's like, well, like again, if I'm a researcher, first of all, I'm going to go where the talent is, where I have like my culture fit, where the money is. And even in places that have a ton of compute and good researchers, Meta still has to offer crazy money to get people to come over. It's like Apple is not going to offer that crazy money. And also, they don't even have compute. And then for inference to serve users, they run it on Mac chips and data centers. It's like very bizarre. And it's like, "I don't want to deal with all that stuff. I want to build the best models," right? Like it's like there's there's it's challenging for Apple. Okay. I I want to ask you one last question about Apple. They are very big on on-device AI, and and I actually really like that approach—security, latency. Um, what's your take on on-device AI, pushing AI to the edge versus having it in the cloud? Is it somewhere in the middle? What do you think?

So, so I think there's—I'm I'm generally an on-device AI uh bear. I don't like it that much. Um, like personally, I think security is awesome. Um, but I know human psychology; like free is better than free with ads is better than security. Um, no one actually cares about security; they say they do, but the number of people who actually make decisions based on security are very little. I would like privacy and security, of course. But wait, but you said you know it's you like free, but you're not—you're not—that's not analogous to on-device AI, right? No, no, no. So, so like Meta will offer in the cloud for free, and OpenAI has a free tier, and you know, so Google has a free tier, and it's going to be better than free, as in running it on your own device. Right. Right. And that's pro—that's a big, big challenge with that is that on-device is that you're you're limited by the hardware. Right. Um, and so how fast the model can inference is really based upon your memory bandwidth of the chip, and oh, okay, if I want to increase the memory bandwidth of the chip, I spend, you know, $50 more dollars of hardware. I pass on the cost to the customer. It's $100 more for the iPhone. Great. With a hundred bucks, I could have like I could have like a hundred million tokens, right? And it's like I'm not consuming 100 million tokens. Or better yet, 100 bucks I just save it, and Meta will give me the model for free on WhatsApp and Instagram, and OpenAI will give it free on ChatGPT, and Google will give it free on Google. Right? It's like it's like it's really challenging from that perspective. And then lastly, um, I don't agree with the latency standpoint, right? I think there's certain use cases where transformers make sense for latency, super tiny uh next-word prediction on your keyboard or a spelling. Um, but the AI workloads that are the most valuable to you and I are—search a restaurant at this time. Yeah. And go find it from a personal standpoint or access to my Gmail, my calendar; that's all in the cloud anyways, right?

Within business, there's tons of use cases. Uh, but my data is all in the cloud anyways.

For personal use, you and I, it's like search restaurant, go through Google Maps, make all these calls, right? Go through my calendar, uh, go through my email. All this data is in the cloud anyways.

A, B, um, if it's more of an agentic workflow, in terms of like, yeah, you know, I'm really feeling Italian and and find a restaurant uh, that's between you and I and location and you know, like we're we're thinking about Italian, but make sure they have gluten-free options cuz he's gluten-free. You know, find me a restaurant with a reservation at 7:00 p.m. tonight. Like this is a deep research query, and then you get a you get a response; it's like, well, that took minutes or like, you know, we we envision the future where the AI books flights for us; it's like this is not a like book the flight okay, it's booked; it's like book the flight; it's researching; it's finding stuff and it comes back, but it's going through the web; it's going through the cloud. Right? Where is the necessity for it to be on device? And because of the hardware constraints, even if it is a streaming tokens thing, your phone cannot run Llama 7B as fast as I can query a server, run Llama 7B and transmit the tokens back to myself, right? And no one wants to run Llama 7B. They want to run, you know, GPT 4.5 or 4.1 or 03 or Claude Opus or whatever, right? They want to use a good model, right? And those models can't possibly run on device.

So, it's a really difficult place for like the use cases there with integrated with all my data, but it's in the cloud anyways. And like it's like how much of my data does Meta have, does Google have, you know, does Microsoft have? Uh, let me plug into all those. Or in the way Anthropic is doing it, they've they've done this MCP stuff and they're connecting in. You can connect your Google Drive to Anthropic, right? And it's like, oh wait, even if I don't have my data with Anthropic, they're still able to connect to it if I give them the rights to. So it's like, where is the benefit of on-device AI truly from a from a use case standpoint? There's certainly one from a security standpoint, but the actual use case is like, yeah, I think there's probably an argument for for a little bit of both. And it probably does skew in terms of like the total workload towards cloud, but I think there's an argument for doing at least a portion of the workload on device, anything that you're interacting with the device on. You mentioned like, you know, typing ahead and that that makes a lot of sense. Yeah, I mean I I I do think AI will make its way on device. I think it'll just be very low-value AI where the cost structure um is just so low, right? I don't think people should design hardware on phones for AI that's going to make it more expensive, right? If you're going to keep the phone the same price point, add AI capabilities, great. But if you're going to increase the price point, I don't think consumers will do it.

How AI on device really will make sense is like, you know, like for example, a wearable, right? An earpiece or a smart glasses. Um, and there you're doing small bits and pieces locally, right? Image recognition, handtracking, but the actual reasoning and thinking is happening in the cloud, right? And that's sort of the the view that sort of like um a lot of these wearables are pushing. I think I think there will be some AI on devices. Obviously, everyone's going to try. It's not like Samsung and Apple and like all these companies are going to sit on their hands. They're going to try stuff. I just think the stuff that's actually going to drive user adoption and revenue and uh improve customers' lives is going to be skewing towards what's on the cloud, which is why Apple has this strategy, right? Apple's building a couple massive data centers, right? And they're they're buying hundreds of thousands of their Mac chips and putting them in data centers. They hired Google's head of rack architecture for the TPU, Andy, um, to make an they're making an accelerator, right? They see cloud as like where AI needs to go. They just like also have to push it on device. But like even Apple themselves, although they won't say it, wants to run a lot of this in the cloud. Yeah. And they do have the they have great chips to do that too.

Um, okay, speaking of chips, let's let's talk about Nvidia versus AMD. I have read a couple articles out of artificial analysis uh um sorry, semi analysis. Um, lately that have kind of uh said that these new AMD chips are actually really strong. Do you think AMD with their new chips h like is that enough to really tackle the CUDA mode or or like are they going to start taking market share from Nvidia?

So I think it's a confluence of things, right? So AMD is um trying really hard. Their hardware is behind in some factors, um, especially uh against Blackwell. Um, but there are some ways their hardware is better, right? Um, and I think the real challenge for them is like you mentioned software, right? The developer experience on AMD is not that great. It's getting better. Um, you know, we we've we've we've asked them to do a lot of things to change it like specific fixes and changes on um CI resources and all these other things. Um, you know, there there's a long list of recommendations. We provided them in December and again more recently. Um, and and they've implemented a number of them um a good number of them, but it's like there's just so they're so far behind on software. It's it's incredible. Now, are they going to gain some share? I think they are going to get some share, right? Um, they had some share last year and they're going to get some share this year. Um, the challen the challenge is like versus Nvidia's Blackwell, it's just objectively worse, right? As a chip. Uh oh. The chip alone, not the ecosystem. The chip alone. Okay. Um, because of the system, right? Uh, because Nvidia is able to connect network their chips together because of the networking hardware they've put on their chip with NVLink, right? Um, so the way that Nvidia can build their servers is like 72 of them work together really tightly, whereas AMD currently they can only have eight of them work together really tightly. And so this is really important for inference and training. Um, and then Nvidia's got this software stack. It's not just CUDA, right? Like people talk about it's just CUDA, but a lot of people don't touch CUDA, right? Most researchers don't touch CUDA. What they do is they like call PyTorch and then PyTorch calls down to CUDA and like automatically like it runs it runs on the hardware, right? Whether it's compile or eager mode. uh whatever you're doing, right? It generally just like maps to Nvidia hardware really well. In the case of AMD, it doesn't as well. And now even less than that, so many people aren't even touching PyTorch, right? They're going to like VLM or SG Lang, which are inference libraries. They're downloading the model weights off of HuggingFace or wherever, right? Um, and they're plugging it into this inference engine, which is an open source repository on on GitHub, either GitHub or uh either SGLANG or VLM. And then they're just saying go and then those things are calling you know torch uh compile and those things are calling CUDA and like you know or Triton and just like there's like all these libraries down the stack. Really the end user just wants to use a model, right? They want tokens; that's it. And Nvidia's building libraries here called Dynamo that make this so much easier for the user. And now obviously there are people like OpenAI's of the world and others who will go all the way down to the bottom, right? Um, you know, DeepSeeks and OpenAI's and Meta's and stuff, but a lot of users just want to call the open source library, tell them, you know, hey, here's my model weights, run. Here's the hardware, run. Right? Um, and here AMD is trying really hard, but it's still a worse user experience. Not that it doesn't work, but it's that like, hey, if I want to use this uh library, it's like for Nvidia, there's 10 flags. For AMD, there's 50 flags, right, that I can, you know, in each of these flags, there's different settings. And it's like, well, what's the best performance? I don't know, right? Like, you know, so so AMD, I think, is is getting there, right? They're getting there really fast, and they're going to get some share.

Um, the other aspect is Nvidia is not doing themselves favors. There's this ecosystem of cloud companies, right? You know, of course, everyone knows about the Google's, Amazon's, um, you know, Microsoft, Azures, right? Those guys have been building AI chips and Nvidia's been trying to which and Nvidia's got AI chips; obviously, they've always been in contention for a while, and so Nvidia as a response propped up all these other cloud companies, CoreWeave and Oracle, not propped up, like really prioritized them. Oracle, but there's actually over 50 cloud companies out there, Nebius and Together and Lambda, and you just go down the list. There's all these different cloud companies um that Nvidia's really helping, right? They're they're taking what would have been allocations to Amazon and Google and others and saying, "Hey, you guys you guys can buy them, right?" Is that is that to kind of level more of a commodity, right? I mean, like you go look at Amazon's margins on GPUs. They're charging like $6 an hour if you were to just rent a GPU without talking to anyone, right? Which is like the cost to buy an Nvidia GPU and deploy it in a data center is like a $140 an hour, right? That's the cost. So then like what's a reasonable amount of profit for the cloud? Maybe $2, maybe $1.75, right? That's what Nvidia wants. They don't want all the profit being sucked up $6 on Amazon. Now obviously you can negotiate with Amazon and get much lower, right? Um, but like you don't want to just like Yeah, it's just it's like really tough. So Nvidia is propping up all these different cloud companies, which is driving down the price. But now they've made a big major misstep in my opinion. They acquired this company called Leptton um who does who doesn't own data centers themselves but they built all the cloud software for reliability for making it run easily, you know, storm, Kubernetes, all this kind of like scheduling stuff. This is stuff the clouds do, right? Which the big clouds do, the neoclouds, the new cloud companies that anybody's propping up do. But now Nvidia's bought this company that does this software layer um and they're doing this thing called DGx Leptton, which is if anyone has a cloud with spare resources um GPUs, just give them to us, and we'll rent them for you, and we'll just give it to us bare metal, no software on them, and we'll add all of this Leptton software on top and rent it out to users. Now the cloud companies are really mad at this because it's like you're direct directly competing with me, right? Um, and in fact, I think Nvidia is also going to put some of their own GPUs on Leptton potentially um that they're they're installing themselves. But it's like you you're you propped us all up, but now you're making a competing cloud. So, a lot of clouds are mad. They won't say it to Nvidia because Nvidia is sort of God, right? You know, like you don't you don't mess with God, right? What Jensen giveth, Jensen taketh. But like they'll tell us the clouds are really bad, right? Um, and so so you know there's this like aspect and so there's some cloud companies that are turning to AMD maybe partially out of AMD paying them, partially out of AMD being invid being them being mad at Nvidia, but like some of these cloud companies are now buying AMD GPUs, and then this there's this third thing that AMD is doing which is they're taking they're doing sort of what everyone accused I don't know if you've seen this CoreWeave Nvidia fraud nonsense, sending revenue back and forth fraud because Nvidia pays them; it's like yeah, Nvidia Nvidia rented one cluster from them. Yeah, cool. Um, seems like business as usual. It doesn't seem like they need these GPUs internally, right? They have to develop their software. I mean, there's like a little something there, but it seems like Yeah, exactly. They invested like a tiny amount of money, right? But whatever. It's like so irrelevant, right? Like um but AMD is actually doing this and like taking it to overdrive. They're getting clusters at Oracle and Amazon and Crusoe and Digital Ocean and TensorWave and they're renting GPUs back from them, right? So, they're selling them GPUs and renting them back. Like, it's one thing if CoreWeave like buys Nvidia GPUs and a small portion of them go to Nvidia, but the vast majority are going to Microsoft for OpenAI. You're not calling this like accounting trickery though, right? It's not accounting trickery. It's like perfectly like the accounting is legal. Obviously, you can sell someone something and then rent something from them. Like Nvidia's done this too. They're almost funding the investment. Right. Right. Right. Exactly. And so this is sort of like in the case of like Oracle and Amazon, it's like, hey, buy our GPUs. We'll rent them back. You'll see that it's great, and you can actually have some some of them you won't we won't rent back. Some of them you'll try and rent to your customers. So that drums up interest, and if it works out, you can buy more, right? This is their this is their reasoning, right? Um, or for the NeoClouds, it's like, well, you guys are only buying Nvidia stuff. Why don't you buy our stuff? Here's here's a contract to get you comfortable, and yeah, here's a portion that you can rent out to other people, right? Like it's like this makes sense to some extent, but it's also to some extent like a lot of the sales are just AMD buying them back, but it's like this fosters really good it's like really good relations right now. TensorWave and Crusoe who are clouds; they're like, I love AMD, right? Because they're renting GPUs for me, and they're selling them to me, and they're renting them back, and I make a profit off of this, and now I can reinvest this in more AMD GPUs, or I have a chunk of AMD GPUs I can rent to other people. Um, and meanwhile, these clouds are like, well, Nvidia is trying to compete with me anyways. Like, what what else am I going to do? So, it's like it's like a it's like an interesting confluence. I think AMD will do well. I don't think they'll like surge in market share, but I think they'll do okay, right? Like, I think they'll they'll sell billions of dollars of chips, but if you're advising a company on which chipset to invest in for the foreseeable future, you're seeing Nvidia. Depends on the price you can get from AMD. I think there's a price where it makes sense to use AMD. And I think AMD will sometimes offer that price to people. Meta uses AMD a good bit. They also use a lot of Nvidia. They use a good bit of AMD. For certain workloads where AMD is actually better, uh, when I have the software talent and AMD is giving me a ridiculous price, yeah, you should do it. And that's why Meta does it, right? But like in a lot of workloads, Meta still goes to Nvidia because that Nvidia is the best.

I want to talk about XAI. I want to talk about Grock 3.5. Obviously, we like at least publicly there's not a ton of information about it. Elon Musk has said it's by far the smartest AI on the planet and it's going to operate on first principles. Uh, is this is this all puffery? Have they actually discovered something new and unique? Uh, specifically that he asked for divisive but true facts, like there there's like a lot of things that he's doing where um it just seems like either he discovered something new or it is pure puffery. What what's your take on what's going on?

I think Elon is a fantastic engineer, engineering manager, but I also think he's a fantastic marketer. Um, I don't know what the new model will look like. I've heard it's good, but you've heard it's good. Everyone's heard it's good, right? So, uh, you know, we'll see what it comes out. U, when Grock 3 came out, I was pleasantly surprised. Absolutely. Um, because I was expecting it to be a little bit worse, but it was actually better than I expected.

Do you use Grock 3 day-to-day? Uh, day-to-day I don't, but there's certain queries I do send to it. Well, like what if you don't mind me asking, uh, their deep research is much faster than OpenAI, so I use that sometimes um and then sometimes like models are just like pansies about like giving me data, right? uh that I want, right? Like it's like you know it's like sometimes I'm just curious like what is the I like human geography like the history of humanity, how geography, politics, history, you know, resources like interact with each other, and so I like to I like to know demographics and things like that as well, right? It's just interesting stuff. You know, the town I grew up in, right, is like it's like on the Bible belt. It's half black, half white, um, 10,000 population. But like the one of the ways I describe it to people is like, well, yeah, it's like where the uh where the flood planes used to be and the ocean receded, it's extremely fertile land. And that's when in Georgia when when when white settlers settled everywhere randomly happened to one of the more fertile areas, so they were able to have better harvest and they were able to then purchase slaves, and that's why it's a higher black percentage than most of the state. And I was like, that's an insane thing to say, but I like to like reason about like sort of human geography like this and like Grock is okay with doing that, right? So sometimes like you know it lets me under and obviously like slavery is bad and like that's you know it's like but like just like to understand like or like hey, why you know oh like invasions from like the step to Europe were not because they just wanted to invade; it's because it was like becoming more arid, and so they had they were like forced off their land and like you know like these sorts of things are cool and interesting or economic history, right? Like why why did Standard Oil win versus this other oil company right before it got to like monopoly levels, right? It's like these sorts of things are just interesting to learn, but like other models will start like if it's like the Standard Oil thing, like it'll be like, "Oh, it was a union buster, blah, blah, blah." It's like, "No, just like tell me like what actually happened, right?" And so, like, I think Grock can sometimes get through the but it's also not the best model. So, my daily the model I go to the most is either um 03 or Claude 4.

Um, you're using 03 day-to-day, even though there's, you know, it takes so much time to actually get your response back. It depends on the topic, but yeah, I think I think a lot of times I'm okay with waiting. A lot of times I'm not. That's why I use Claude. I use Gemini in in work, right? So, we feed a lot of permits and regulatory filings through Gemini. We feed a lot of like u long it's like really good at long context, right? And document uh analysis and retrieval. So, we feed a lot of stuff through uh Gemini in a workplace manner, but like I'm talking about pull out my phone, I want to know something mid-con conversation or whatever. It's a different model. So, okay. So, so back to So, Grock, yeah, Grock, they they have a lot of compute. It's really concentrated. They have a lot of great researchers. Will they, you know, they've got like 200,000 GPUs already up. Um, and they've purchased a new factory in Memphis and they're building out a new data center and they're shi, you know, you know, there's the craziness they did with like mobile generators. Well, now they just bought a a power plant from overseas and are shipping it to the US because they couldn't get a power plant uh, you know, a new one in time. So, like they're doing all this crazy to get the compute. They've got good researchers. Clearly, the models are good, and Elon's hyping them up. Maybe it'll be great. Maybe it'll be good. Maybe, you know, will it be OpenAI level or it'll be like just slightly behind, you know, I don't know. But are they doing something fundamentally different? He specifically said rewrite the corpus of human knowledge because too much garbage is in the current foundation models that I mean I mean he must e obviously he has the x data which is insane but it's also really low quality so it's hard to get.

Through. Right. So, but also at the same time, uh oh, that's another area where I use Grock sometimes: current events. Um, yeah, summarizing or giving you the contiles are happening in Israel and Iran and all this war stuff. You can actually ask Grock, and it tells you what exactly what's happening way better than uh a Google search will, or even a Gemini query or open query, because it it's got access to all this info. Yeah.

So, are they doing anything different? I I like like step function different, like we're I think step function different-wise, I don't think anyone is like, you know, like everyone likes to think they're doing different things, but generally people are doing the same thing. They're pre-training large transformers, and they're doing RL on top mostly in verifiable domains, although they're researching how to do unverifiable domains. It's like it's like it's like oh yeah, they're making environments to to for the model to play in, but they're mostly code and math, but now they're getting into like computer use and, you know, all these other things. It's like everyone's doing generally the same stuff, but there's such a it's also like such a challenging problem. There's many directions to go with it, but I think generally everyone's doing going the same approach. Even SSI is not I I imagine SSI is doing some different stuff, but I don't even think they're doing that much differently than than what I just said.

I have kind of two different topics I want to let you maybe choose. Uh, economics, labor. So, I want to talk about the 50% of white collar jobs could disappear. I know you probably read about that, or um non-verifiable rewards, which um, you know, obviously that's maybe more recent, more on your mind. You have any preference? Um, I think the prior is more I mean maybe maybe the latter is more interesting for your your uh your your audience. Uh, I'm not sure, but the prior is really interesting, right, in terms of like everyone's worried about massive job loss, right, um or at least some people are uh in the AI world, but then the flip side is that like, you know, populations are aging really rapidly, and generally people work less than ever before, right, like we make fun of Europeans because they work really a lot less, but like the average amount of hours worked 50 years ago was way higher, right, and 100 years ago is even higher than that, and the amount of leisure time was way less, and the size of everyone's home is way larger, and like the food security is way better. It's like every metric were way better than 50 years ago or 100 years ago, and AI should just enable us to work even less right now. Is it going to be like there's going to be psychos like like myself and probably yourself as well that work way too much. Um, and then there's going to be like normal people who work way less, right? And and you know, obviously the distribution of resources is the challenge though, right? Um, I think that's the big thing.

Um, that's why I'm super excited about robotics as well, because robotics is like uh, you know, a lot of jobs that are easier to automate are are hardest to automate are robotics influenced, and the stuff people want to do is sit on a computer and be creative, but actually that's one of the markets that's been nuked the hardest is freelance graphics designers, right? And what's the market that's like not touched is like picking fruit, right? And it's like it's like that's the that people don't want to do. And it still seems like that's way in the future, even though robotics has been progressing at an in an insane rate, but it does seem like it's pretty far in the future still. Okay.

But um, do you foresee as human productivity increases like crazy? Uh, certainly a a large swath of tasks will be automated um, do you think humans are going to be managing AI in the future or are we going to be reviewing the output of AI or some mixture of in between? Right now we're in the transition from using models on a chat basis to a longer horizon basis. You know, I mentioned I used 03 a lot because actually there's a lot of longer horizon tasks. Now these longer horizon tasks are 20 30 seconds, and deep research is dozen minutes, right, dozens of minutes. Um, over time these like interactions with AI will become obviously there will be an AI assistant that I'm just talking to all the time or will be telling me stuff that is not noteworthy, but there will also be just long horizon tasks of like AI is going to be doing stuff for hours, days before coming back for me to review, and then eventually like there just won't be humans in the loop, right?

Um, and eventually do you believe that? And like what timeline are you thinking? Uh, I think timeline questions are ridiculous. I'm I'm generally more pessimistic on timelines. It's not I don't think this decade for people to you know for like 20% of jobs to be automated, I think it's like like not it's like maybe the end of this decade, maybe the beginning of next for 20% of jobs to be automated, right, meanwhile there's people saying AGI in 2027, u but their definition but reaching tech doesn't mean the implementation is going to happen at that moment either, right, it's going to take years before we actually are are able to deploy it in the field. I think I think deployment will be really fast, you already see the junior software engineering market is nuked, no one can get a You can already see um the usage of AI and software development is skyrocketing. Um, and we're not even at automated software development yet. We're just at like code assistant. Are companies going to choose to do more things? Are they going to choose to tackle more problems? Yes. So then like how do those junior engineers get into the market to begin with? Then I assume like I spoke to Aaron Levy yesterday and he was like no, as soon as a team tells me look how productive we are. Where do you think I'm going to invest? I'm going to invest back in that team. We're going to grow that team. um, where is the place for junior engineers then?

Yeah, I think I think that's nice and I agree and like I myself like my company does a bunch of stuff, but but due to the use of AI we can do a whole lot more stuff, and that makes us more productive, and we're able to out compete the old firms that don't do stuff uh in the consulting and data space. Um, but it's like, you know, I still I still have like basically doubled the size of the firm in the last year to 32 now 33. How many junior software developers am I going to hire? It's like no, it's like the junior software developer I have like we just cheered her on because she just did like 50 commits like last week, and it's like that's what used to take many more people. It's like how much stuff like there's obviously a lot of software for us to build. Um, but it's like, you know, how many people can we like actually like uh, you know, add right, and it's like wouldn't I rather have like a senior person that's like commanding a bunch of AIs rather than like a junior person. So it's like sort of like it's challenging u at the same time like, you know, hiring young people because they can't they do they can quickly adapt to the new AI tools, right, sort of like it's a it's a it's a a balancing act. Um, I think it's I think it's I don't know where the junior software developers would go because I get people pinging me on like Twitter and LinkedIn all the time like you have a job for me. It's like no, I don't really, but like uh um or sometimes I do, right? But you know it's like it's tough, and I don't see the major tech companies hiring junior software developers that much, right? Um, it's just a fact, right? And that's why the market is really bad. Um, so they had to just self-skill up on their own. Come come come with better skills. But or try and build stuff on their own and show that they're not a junior software developer, but they can actually use these tools. That's not for everybody though. Yeah, it's not it's not People A lot of people just need a job. They don't need to like self-start and like they don't want to be founders for sure. They don't want to be kind of solo builders. Even if you're not a founder, they want to have that. Yeah. I mean, that's been a problem for me is like when I started hiring people is like some people need a lot of direction, and I don't have direction to give. I'm like, I need self-starters. Right now there's like people who can you know do that like in the firm, but like it's like it's tough to like give people you know people some people just need direction and need more hand-holding at least initially.

Open source versus closed source and why the US is going to lose in open source um unless Meta gets dramatically better, which they are. I think with a lot of the talent they're hiring. I think Sam is wrong that they're not getting any top researchers. Um, Sam Altman. I think they are um there's some top researchers I know for sure are going there. um maybe not the first people they offered, right? Like the ones that like have the highest highest profile, but there are still some top researchers going there. China is open sourcing stuff only because they're behind. The moment they're ahead, they will stop open sourcing stuff. And at the end of the day, closed source will win. Um, unfortunately, closed source will win. My only hope is that it's not just like two or three closed source AIs that like dominate human GDP, right? Or or types of models or companies, right? But rather it's like more distributed than that. But it might not be right.

Meta, Google, OpenAI, Microsoft, Tesla, whoever else, you had to pick one company to bet on. Super intelligence reaching it first. Who are you picking and why? OpenAI. Um, they're the first to every major breakthrough. Um, even reasoning, they were the first two. Um, and I don't think reasoning alone will take us to the next generation. So, there's going to be something else. Anthropic second. Um, third would it's a toss-up. They're so conservative though. They're so conservative at Anthropic in terms of what they release, what they publish, what they focus on. So much safety. I mean, I think has weakened a lot. I think they're a lot less conservative than they used to be. Okay. Um, like the process for launching Cloud 4, as far as I understand, was much simpler and easier than the process for launching Cloud 3. Whether it's that they're hiring a lot more normies, which they are. um or like they recognize that others are just going to release stuff anyways and they should have theirs or whatever it is. I think Anthropic is like loosening uh up a bit. I think they just have really good people though. Um, and then sort of like third is going to be like it's actually a toss-up between Google XAI and and um X and and Meta now. I think Meta will get enough good people that they'll actually be competitive too.

Dylan, thank you so much for chatting with me. Thanks for having me. Appreciate it. This is awesome, man. Yeah, very fun. Yeah, you can uh talk about anything, huh? Maybe. Maybe.