A (2:32)

Been like that for years.

B (2:33)

It's like, yeah, no, you've been training to be, you know, to have terrible airpod takes literally since you came out of the crib.

A (2:41)

That's right. Yep.

B (2:42)

Anyhow, what I appreciated about Nico is, is there's a bit where it's easy to dismiss someone like that, say, oh, well, they had all these advantages in life. And what is striking is how he doesn't deny he's had advantages. He just relentlessly identifies his advantages and then leverages them to do the next step. Right. And that's a bit. That's what distinguishes people who, sure, they start out in a good place, but not everyone who starts out in a good place actually accomplishes big things because there's a bit about you. You have to actually take advantage of your circumstances. And I just thought that, like, it was interesting to tie that all together. It was just a little bit of, like, you know, is this like a hardcore, like, I've been talking to, you know, there's a Jensen Huang interview in the news this week. I might write about it next week. Now, this is a bit of a softball sort of bit, but it was actually was more interesting to me than I expected from, for that reason, about, like, this linkage that you wouldn't think exists, but does 100%.

A (3:45)

And what I appreciate Nico for is his uncomfortably frank F1 broadcasting and his questions to F1 drivers. You can tell that he doesn't need that job.

B (3:57)

Exactly.

A (3:57)

He has fun with it and is going to speak his mind the entire time. So shout out to him.

B (4:03)

I was disappointed.

A (4:04)

Welcome as the third chair on this podcast anytime he wants.

B (4:07)

Well, I was disappointed. He, he, he was not as critical of the current F1 regulations.

A (4:12)

Mercedes is cutting him a check.

B (4:14)

After all, he is still associated with

A (4:16)

Team Mercedes, so gotta play it diplomatic for now. As for the show, Ben, we are gonna start with a six pack of AI questions that kind of run the gamut here, just bouncing all over the place. Daniel, we'll begin with Daniel writes, longtime reader and listener here, about Monday's article from Ben on AI and aggregation theory. Isn't the central question, which I felt went unaddressed, whether having the Best product in the consumer space specifically is going to go from zero marginal cost to high marginal costs. In the latter scenario, the products that consumers will expect in the future are thinking and agentic and would therefore always benefit from more computer. That means the primary unit of quality goes from upfront design and implementation to continuous build out of ever more capacity to serve ever rising per user token demand. What do you think of that question?

B (5:14)

It's a good question. It puts its finger on something that is somewhat unresolved. I try to put forward that it was unresolved, sort of. Particularly at the end, which is, oh hey, let's assume best product wins. But yeah, I'm just repeating Daniel's point. What if best product means most compute and that's actually what matters. So I think there's a few problems with the premise of this question, which is, or the premise of this assertion. So number one, it sort of assumes that these agents will never be good enough, that there's basically an infinite like capacity for, for, for how quality they could be. And that might be the case. I did write an article about sort of like, you know, one of the brilliant insights of building a product that's focused on consumer satisfaction. I think it was talking about this in the context of Amazon is no one is ever actually fully satisfied. So it's like, it's like it's a goal that you can never reach, can

A (6:13)

always be more convenient. Sure.

B (6:15)

And so this sort of fits in that framing, which is can actually agents always be smarter to the, you know, to the extent that the sort of improvements can continually be absorbed. Just like the improvements in like the user experience I've put forth previously can be continually absorbed. That might be the case, but it's,

A (6:37)

it might be the case, but as a normie, can I just interject to say that the differences between models are already very difficult to perceive. Now as a user, and if we're talking about the consumer space specifically, which is what he's focused on, I'm not convinced that the performance delta is going to be what decides who wins in that space over the long term because everything is so performant already and it's only going to become more so over the next couple of years.

B (7:04)

I know, I, I completely agree. And like, implicit in this, this is actually a very optimistic framing in terms of like the doomer narrative around AI because you're basically stating AI is never going to be smart enough for humans, so it could continually get smarter and smarter we, which maybe that's the case, but it's worth Pointing out this is contrary to the narrative that AI is going to surpass human capabilities. To the extent it surpasses human capabilities, then it is going to be good enough. Which means the increase in quality and capability to your point, is not gonna be noticed or appreciated because it's already good enough. In which case we are reverting back to a zero marginal cost world where the computing is effectively free. Because o oh yeah, oh, the AI is up there in their planning meetings and saying, oh yeah, the humans can run on the model from 13 years ago because they don't know any difference. It's good enough. Right. And running on like, you know the, oh, those chips from 2035, those don't matter. Like super. Like that's like which one is it? Like there's. So there's a bit. There's this overall attitude towards AI of always choosing the pessimistic interpretation of everything, even when they're totally in conflict with each other. Other. And this is a good example. This insistence that compute is going to be consumed infinitely is in direct conflict with the idea that AI is going to get smarter than humans. Once it's smarter than humans, it's smarter than humans. Right. At least to your point. If we're talking about the consumer market.

A (8:38)

Yeah.

B (8:38)

And you made the point, what do consumers actually want? Like, if it's just an answer bot, we're already in a pretty good position. To your point. Right. We sort of saturated the space remarkably quickly now. I think there's a bit where one reason it makes sense to focus on enterprises right now. AI is clearly a productivity booster and I've been making this point repeatedly. Enterprises pay for productivity and consumers don't. Consumers don't want to be productive. They want to enjoy themselves and have fun. Like watching reels is not productive, but it is enjoyable.

A (9:14)

It's a great business, right? Yeah.

B (9:15)

And it's a very good business. Right now, of course there's this idea of we're going to have assistants that help us do things and whatever it might be. But at the end of the day, a lot of stuff, the point is the friction. Like what do people do for fun? Actually, I saw, I think I saw a reel. I'm going to like badly interpret a comedy bit which is rich people. The way to identify an activity that rich people enjoy doing is it's a reinterpretation of what poor people already do. So like, for example, let's go hiking, walk very long distances and like tire ourselves out and admire that. Guess what? You could walk as if we don't have time. That's right. Right. There was a whole list of things I'm not going to like, reinterpret it. I think some of them were fairly objectionable, as any good comedy bitch should be. But like there is an aspect here of this is maybe a reason why enterprise markets are just always different than consumer markets. And maybe a mistake companies make again and again and again is for getting this distinction. And there's that kind of gets into my bit. It probably makes sense for OpenAI to double down on enterprise because that's where the money is for the next five years AI use case which is increasing productivity. That though is a reason to be quite optimistic about Meta because they're not in the enterprise space and they already have a functioning advertising business more than functioning. And OpenAI just like they can't do both and it's probably a mistake to try to do both. And sure, ChatGPT is still the biggest today, but every little bit of time

A (11:08)

they can't do both. Are we basically talking about optimizing the ads engine on the consumer side to monetize that over the next couple of years while also focusing on the enterprise side and dedicating compute to the enterprise side? Because I don't really see why they can't do both.

B (11:27)

Well, I mean it depends on, you know, I guess you're the AI optimist here. They could just have AI write all the code to build up their ad engine on the consumer side and whatever lesson pill.

A (11:38)

You know what I mean?

B (11:39)

I mean maybe you're right. My general view. Yeah. Is still based in a view of talent scarcity. Like actually the building that out is going to take a lot of time and focus and energy that I see. You should probably be building out more of a business in a place that

A (11:56)

actually pays hard to win in two worlds that are ultra competitive. And Meta already has very different.

B (12:04)

Yes.

A (12:04)

Pretty well developed.

B (12:05)

Well at Google as well. I mean I don't mean to dismiss Google in this sort of story here. The Google story is pretty interesting just because they were the like December wasn't that long ago and I was like, oh, Google's one, it's all over.

A (12:21)

Well, you know I actually on the six pack of questions here, I have this question from Robin who says, do you think the future narrative for this year might be Gemini falling behind by not milking the compute scaling law via Blackwell? Because I do. And I included that note from Robin mainly because it was the first time in months that someone has mentioned Gemini in our email so do you have an answer for Robin and what do we make of good old friendly Google right now?

B (12:50)

I don't know. I mean like the like is Google fundamentally hindered by not using Nvidia? I don't know that that's the case. You know, it, you know, it's hard to know any of this stuff for sure. The sort of feedback I generally get and I'm not an expert in this space, so this is mostly sort of second, third hand is Gemini's not very good at coding. And just in general like the. It's interesting because Gemini feels like a model that, you know, Facebook got a lot of grief and I think had a very sort of traumatic event to the extent they basically fired an entire team, fired everyone because they were baked. They were, they were, they cheated on benchmarks by and basically like and Gemini, I don't know that it was done intentionally. It feels like a very benchmark optimized model that people actually use and find it fairly unsatisfying to use. Now again, I'm not necessarily the right person to judge this other than. And I think Gemini would say, look, in real world use cases in the enterprise, people are using Gemini up and down, left and right. And yeah, we're, you know, just because we're not releasing half baked products and getting, you know, on the hype cycle doesn't mean it's not a real thing. And going forward we have this integrated advantage. We have all this compute, et cetera, et cetera. And by the way, it's worth noting, Google cloud numbers are awesome. Right? Like, so the business results are there. Is that Gemini, is that because they're renting out the cloud to other folks? I mean, who can tell for sure? But, but yeah, like your takeability index almost needs just a separate section for AI specifically because anthropic is clearly top of the world right now. But is it going to be the case that in a month, like is the narrative gonna be totally different and totally flipped on its head? It sure seems like that's possible.

A (14:53)

Possible, right. And a company like Google, it's like, well, what can you say after the last couple months, all the advantages that we talked about in December are still there and they're just not necessarily winning the hype cycle every couple of weeks here. But by the same token, like they have the funnel to consumers with Google and they have gobs and gobs of money to continue throwing at the problem. And on an infrastructure basis, they're well positioned to own this space over the next 10 years. Here so it's hard to get like too down on them. But it's also kind of curious that they're just never really mentioned in these conversations.

B (15:29)

I don't really know anyone that uses them for coding. Like that seems to be a real weak point as far as far as development goes, and we've talked about that.

A (15:37)

And coding is what's driving all the hype right now.

B (15:39)

That's right.

A (15:40)

So Google's sitting that game out. All right.

B (15:43)

Well, I don't think there's anything. There's certainly, you know, but well, they're

A (15:46)

losing, they're on the sidelines, they're benched. Thomas says. Ben and Andrew, I was thinking about the current conversation around Anthropic and OpenAI and in particular how their beliefs about the path to AGI seem to be shaping their current positions. Anthropic is extremely strong at coding but constrained by limited compute, while OpenAI seems to have more reliable access to compute, yet may be slightly behind in coding performance and adoption. I think it all comes back to OpenAI having always been the scaling laws company and Anthropic focusing on recursive self improvement. Dario talked about why he didn't buy more compute on Dark Quest or Dwarkesh from an economic standpoint, but I also believe that he sees the path to AGI as the recursive self improvement from the AI being able to train the next AI. When you believe this, it doesn't make sense to waste time on image or video models that just take away resources from the specific type of model needed to build AGI. It also means you can't overspend on compute because you need to make sure you hit the takeoff just right. OpenAI is and always has been about scaling laws, and if scale is all that matters, then you can't possibly overspend on compute. Codex was a late focus because they saw the success of Claude code, but they have always been the AI company that takes scaling laws literally. And if you believe in scaling laws, there's no such thing as too much computer. Both companies are learning from the other in these scenarios leading to larger compute spend by Anthropic and more focus on coding by OpenAI. But I think the winner will be determined by how you can get to AGI. So Ben grade the theory there. What do you think of that breakdown?

B (17:32)

I think it's a pretty good one. I think one way to think about why Anthropic has been doing so well and OpenAI was scuffling a bit I think does come down to some Combination of focus and alignment. The anthropic is super focused on coding, as Thomas notes, because it sees coding as the way to. Once the AI can program itself, then yes, recursive self improvement. And that's the actual sort of takeoff. And humans aren't going to program the AI to AGI. The AI is going to level itself up into AGI.

A (18:10)

So is it correct to say that anthropic is the most bitter lesson pilled of the Frontier Labs?

B (18:15)

No, this is, I think this is kind of the opposite of bitter lesson pilled. Like they think they can like for them algorithms matter. The thing is you just have to get to the algorithms, writing their own algorithms. But yeah, it's interesting to put that bitter lesson sort of framework. My initial response would be no, it's actually slightly different. And so the reason why that is great for them from a business perspective is as we've discussed repeatedly, coding is a great application for AI. You're generating a lot of text and it's verifiable which solves the problems of coming up with a bunch of texts that might be hallucinating or it's possible to systemically find issues and mistakes and go back and fix it and get this sort of recursive loop going on. And but this is very powerful from a business perspective because they are motivated by building God, as it were. Their way to build God is turns out to be a product that everyone wants to buy just to make business applications. Right. And so everyone's pulling in the same direction. OpenAI has had more, more of a challenge. OpenAI to your point when he's talking about scaling, he's talking about bitter lesson is about, is about scaling laws. It's just get more data, make it bigger, do more and more and like more compute and more data is going to solve all your problems for you. And so I think that, yeah I think is that OpenAI sort of overarching concept maybe. I think that's not, not, not a bad way to put it. You also have a bit where what OpenAI's biggest advantage is like they're really good at this reinforcement wording and reasoning. Like OpenAI's models today are still significantly smaller. They're more like GPT4 class models than like say a Gemini, which is much larger. But the reason why it feels better is if you use the plain chatgpt it's terrible. You could, you could easily tell it's not as good you like but where it's good is if you're in like thinking or in Pro where like, the reasoning is just really good. Like it just. And it's. It's doing multiple things, it's comparing them which one's better and like, like it just. It's also makes it very slow. And so there's like. One of the things with ChatGPT is maybe this is another reason why the consumer might end up not working out for them, is because the best consumer models, I suspect, are actually larger models that just turn out close to a right answer the first time.

A (20:45)

Decent answers, right?

B (20:46)

Like sitting around for ChatGPT. Like, if you want to get the best out of ChatGPT, you have to use the modes that are super slow and you have to sit around and wait for it. And it's kind of a crappy experience, honestly. But it does give you really good answers. And, you know, there's a lot of people that, you know, actually think their coding capabilities are. Are much better than quads in many respects. But a lot of this is downstream from. They're just really good at reasoning and it's slow and it take. Takes a long time, but that's a misalignment. It's not aligned necessarily with what they're doing. They've done like 47 different things, no focus. They have a. So they have a research team that also wants to get AGI. They see as reasoning as a way to get there that's not necessarily aligned with what they're trying to build from a business perspective.

A (21:29)

And so then the scaling concept, I mean, because those are two totally different things.

B (21:34)

Yeah, well, I mean, I think their new model, you know, I think they're going to have a new class model like this Spud model. It's going to make Spud unique is it's like truly the next generation. It's going to. And I actually think it's gonna be really interesting to see how SPUD is, because if they can have a much better base model with their sort of RL layer and their reasoning capabilities, like, it could definitely be very, very capable. And so maybe that's gonna drive like the next change in narrative or sort of whatever.

A (22:01)

I was gonna say continually updating the takeability rankings in the AI space. Who knows where OpenAI will be in three weeks.

B (22:08)

But I do think what Thomas definitely is right on is Anthropic's focus on coding has been hugely beneficial. And if OpenAI just internal alignment has been a challenge for OpenAI, that's been all the upheaval inside the team and Anthropic sort of gotten that for free. And that's a huge factor in their success.

A (22:29)

Fair enough. Okay, well, a different Thomas wrote in and said several times Ben has complained that TSMC left money on the table when they had the best process. But were production constrained, surely they should just raise prices until the demand balanced. Well, likely they worried that some of that balance would come from customers getting used to Samsung's process or even Intel's. And those losses could be sticky in the AI world. If we're now compute constrained, should the AI companies be raising rates? What do you think? All right. And that is the end of the free preview. If you'd like to hear more from Ben and I, there are links to subscribe in the Show Notes or you can also go to SharpTech FM. Either option will get you access to a personalized feed that has all the shows we do every week, plus lots more great content from strikeri and the strikeri Bundle. Check it out and if you've got feedback, please email us at emailarptech fm.