OpenAI Misses Targets, Codex vs Claude, Elon vs Sam Trial, Big Hyperscaler Beats, Peptide Craze

Chamath Palihapitiya: Everything in this market is power constrained. The reason that these folks may miss a number or a forecast have nothing to do with demand. It is entirely 100% due to the supply of the power necessary to generate the output token. There was a really interesting thing that was just announced today that will make this problem even worse, which is what you're starting to see now is backlogs build up of not just the access to the power, but then the componentry that's actually necessary, not just recips and not just nat gas turbines, but now you're talking about transformers and all the actual tactical grid infrastructure. Why is this important? If you look at the actual amount of gigawatts that are under construction, we have a huge mismatch now. People have announced all these projects, Jason, but less than half of it is actually being built. Less than half. Most of it is stuck in red tape. Most of that is because there are these supply chain delays. So there's no credible strategy to turn any of this stuff on. Who will this hurt? It will hurt Anthropic and OpenAI the most. Who will this benefit? It will benefit the hyperscalers, specifically Oracle, Amazon, Meta, Microsoft, and Google.

Chamath Palihapitiya: Everything in this market is power constrained. The reason that these folks may miss a number or a forecast have nothing to do with demand. It is entirely 100% due to the supply of the power necessary to generate the output token. There was a really interesting thing that was just announced today that will make this problem even worse, which is what you're starting to see now is backlogs build up of not just the access to the power, but then the componentry that's actually necessary, not just recips and not just nat gas turbines, but now you're talking about transformers and all the actual tactical grid infrastructure. Why is this important? If you look at the actual amount of gigawatts that are under construction, we have a huge mismatch now. People have announced all these projects, Jason, but less than half of it is actually being built. Less than half. Most of it is stuck in red tape. Most of that is because there are these supply chain delays. So there's no credible strategy to turn any of this stuff on.

David Friedberg: And I just want to highlight this paper that came out from MIT from these two scientists. And these guys published a paper on pruning techniques in neural networks. This paper showed that you could actually reduce the size of these networks by 90% and get the same accuracy out by pruning very large models down to smaller models, and then you can make a selection on which model to run for inference. And by doing this, you can actually reduce inference costs by 10x. You can get 10x the output per energy unit that goes into the data center with no loss of accuracy. And so it's a really interesting, call it algorithmic technique that can be applied to the existing large models to actually make them much lower energy use. So if you think about it, you're firing up a very large model to answer a very simple question. You can actually prune away that model. Now, this is probably going to be the case in AI applications as it is in traditional Google search. There's a long tail of searches, but there's a few searches that account for a large percentage of search volume. It's like, what is the weather? What are the movies, times? What's the stock price? Like there's a certain set of things that make up the bulk of consumer energy. And there's probably a certain set of things that probably make up the bulk of coding output as well. And so if you can get that 80% of searches or chat interfaces or coding requests, reduce down through pruning techniques to smaller models, and then you have a whole set of smaller models that can be called dynamically, and you reduce inference cost by 90%, you can make much more use, call it 10 times the use on data center and energy capacity than we can today. So I would argue that we're still in the very early days of getting efficiency in terms of output and tokens, and we're just in the very kind of early stage of that, which also unlocks the opportunity for guys like Elon to reinvent how this is done and potentially compete pretty aggressively.

David Friedberg: I just want to highlight this paper that came out from MIT from these two scientists. And these guys published a paper on pruning techniques in neural networks. This paper showed that you could actually reduce the size of these networks by 90% and get the same accuracy out by pruning very large models down to smaller models, and then you can make a selection on which model to run for inference. And by doing this, you can actually reduce inference costs by 10x. You can get 10x the output per energy unit that goes into the data center with no loss of accuracy. And so it's a really interesting, call it algorithmic technique that can be applied to the existing large models to actually make them much lower energy use. So if you think about it, you're firing up a very large model to answer a very simple question. You can actually prune away that model. Now, this is probably going to be the case in AI applications as it is in traditional Google search. There's a long tail of searches, but there's a few searches that account for a large percentage of search volume. It's like, what is the weather? What are the movies, times? What's the stock price? Like there's a certain set of things that make up the bulk of consumer energy. And there's probably a certain set of things that probably make up the bulk of coding output as well. And so if you can get that 80% of searches or chat interfaces or coding requests, reduce down through pruning techniques to smaller models, and then you have a whole set of smaller models that can be called dynamically, and you reduce inference cost by 90%, you can make much more use, call it 10 times the use on data center and energy capacity than we can today. So I would argue that we're still in the very early days of getting efficiency in terms of output and tokens, and we're just in the very kind of early stage of that.

David Friedberg: this paper came out from MIT from these two scientists. And these guys published a paper on pruning techniques in neural networks. This paper showed that you could actually reduce the size of these networks by 90% and get the same accuracy out by pruning very large models down to smaller models, and then you can make a selection on which model to run for inference. And by doing this, you can actually reduce inference costs by 10x. You can get 10x the output per energy unit that goes into the data center with no loss of accuracy. And so it's a really interesting, call it algorithmic technique that can be applied to the existing large models to actually make them much lower energy use. So if you think about it, you're firing up a very large model to answer a very simple question. You can actually prune away that model.

David Sacks: Because look, there is so much fear right now, almost the level of panic about mythos. People are treating it like a doomsday weapon or something like that. It's not. It's simply that the frontier models have reached the point where they're capable of automating cyber activities, just like they're capable of automating coding. But that means that a model could power up a cyber attacker or cyber defender the same way they can power up a coder and allow them to discover a lot more vulnerabilities. So there is obviously a risk there, but I think it's important to understand that Mythos or GPT-5.5, it doesn't create the vulnerabilities. It just discovers them. The bugs were already in the code. They were sitting there waiting for some hacker to discover. If we can now use AI to find these bugs in advance, these vulnerabilities, and patch them, then you actually harden our infrastructure and you harden our security. I also believe that this leap from, let's call it pre-AI cyber to post-AI cyber, it's going to be, I think, a big one-time upgrade cycle, because again, you're going to find all these dormant bugs and vulnerabilities. But I think that once we get past that upgrade cycle, you're going to reach a new equilibrium between AI-powered cyber offense and AI-powered cyber defense. It's going to become a lot more normal. It's not going to feel like this huge disruption, which is to say, I think people are treating this as like some existential threat. I don't think it is as long as everyone does what they're supposed to do, which is use the new capabilities to harden their code bases and infrastructure and security before the hackers get ahold of these capabilities.

David Sacks: there is so much fear right now, almost the level of panic about mythos. People are treating it like a doomsday weapon or something like that. It's not. It's simply that the frontier models have reached the point where they're capable of automating cyber activities, just like they're capable of automating coding. But that means that a model could power up a cyber attacker or cyber defender the same way they can power up a coder and allow them to discover a lot more vulnerabilities. So there is obviously a risk there, but I think it's important to understand that Mythos or GPT-5.5, it doesn't create the vulnerabilities. It just discovers them. The bugs were already in the code. They were sitting there waiting for some hacker to discover. If we can now use AI to find these bugs in advance, these vulnerabilities, and patch them, then you actually harden our infrastructure and you harden our security.

Chamath Palihapitiya: The reason that this is even possible is because humans are error prone. And when humans code, they create holes. And so humans exploiting humans is where we've been for a long time. Now we have computers exploiting humans because the computers go and seek out all these bugs that humans wrote. In the next phase, it'll be machines versus machines. And so I think the nature of cyber is going to completely change. Probably in the next five or six years, there'll be so much reason to rewrite all of the software that runs the world. In one part because you're going to be asked to show more operating leverage and revenue growth, but in another part because everything else that was handmade in the past is just fundamentally insecure. Either way, all roads will lead to all the operational software that runs the world will get rewritten. More and more of it will be written by machines. More and more of it will be impregnable as a result. But then the cyber threat actually will only increase. Because then you're going to try to figure out how to use a machine to inject something into another machine so that some agentic loop injects some malware or injects a bad token.

Chamath Palihapitiya: The reason that this is even possible is because humans are error prone. And when humans code, they create holes. And so humans exploiting humans is where we've been for a long time. Now we have computers exploiting humans because the computers go and seek out all these bugs that humans wrote. In the next phase, it'll be machines versus machines. And so I think the nature of cyber is going to completely change. Probably in the next five or six years, there'll be so much reason to rewrite all of the software that runs the world. In one part because you're going to be asked to show more operating leverage and revenue growth, but in another part because everything else that was handmade in the past is just fundamentally insecure. Either way, all roads will lead to all the operational software that runs the world will get rewritten. More and more of it will be written by machines. More and more of it will be impregnable as a result. But then the cyber threat actually will only increase. Because then you're going to try to figure out how to use a machine to inject something into another machine so that some agentic loop injects some malware or injects a bad token. And I think that's a very complicated thing. What I will tell you is, I'm not even sure if I'm allowed to say this, but a very good, probably the best cybersecurity company in the world, run by one of the very best CEOs in the world, who may or may not be speaking at Liquidity, would tell you that they have penetrated and can essentially manipulate every model.

Chamath Palihapitiya: the nature of cyber is going to completely change. Probably in the next five or six years, there'll be so much reason to rewrite all of the software that runs the world. in one part because you're going to be asked to show more operating leverage and revenue growth, but in another part because everything else that was handmade in the past is just fundamentally insecure. Either way, all roads will lead to all the operational software that runs the world will get rewritten. More and more of it will be written by machines. More and more of it will be impregnable as a result. But then the cyber threat actually will only increase.

Jason Calacanis: Elon is, of course, accusing OpenAI of breach of charitable trust, unjust enrichment. He's accusing OpenAI of essentially flipping A non-profit into a for-profit. He's seeking $150 billion in damages that they revert back to a non-profit, that Altman and Brockman be removed. And there were some fireworks between Elon and the OpenAI lawyers. Elon kind of leveled up the discussion. He said, Quote, if we make it okay to loot a charity, the entire foundation of charitable giving in America will be destroyed. That's my concern. Obviously, there's a ton of interesting nuances here. Specifically, Greg Brockman keeping a diary where he was journal maxing his plans, like a Bond villain here. And the excerpts from his diary include, conclusion, we truly want the B Corp. The true answer is that we want Elon out. If 3 months later we're doing B Corp, then it was a lie. Can't see us turning this into a for-profit without a nasty fight. I'm just thinking about the office and we're in the office and this story will correctly be that we weren't honest with him. In the end, it's still about wanting a for-profit just without him, yada, yada, yada.

Jason Calacanis: I just don't know why Greg Brockman's got a frigging diary where he's like literally documenting. I mean, I love the guy, but what the fuck is he thinking? Like, you're just sitting here at home and like, let me write about the crime I'm committing or let me write it like, and let me record it. And by the way, let me never delete it. I don't understand this. It's not just journal maxing, it's discovery maxing. It's smoking gun maxing.