Haseeb

And there's also been a lot of talk lately about regressions within Anthropic.

So if you look at the uptime chart for Anthropic, it's pretty bad.

Yeah, it's very clear that they are falling over from all the demand.

And so they're just not able to scale their GPU capacity for all of the demand that they're getting.

And it's only accelerating from here.

So I would expect that this might be the opening for OpenAI is that they do have access to more compute.

They've been much more aggressive in securing GPU demand, sorry, GPU supply.

And so that may mean that

If Anthropic just keeps kind of falling over and not being able to solve their supply problems, that the stability of open AI allows them to start catching up and we end up back in a kind of duopoly situation.

Yeah, I think a few things.

I spoke with Justin Drake about this recently.

I think the multi-client architecture for Ethereum will probably go away in a post-AI world.

And I think a lot of the reason for this is because one, Ethereum is going to need to be formally verified.

Yeah.

The defense when software becomes incredibly cheap against potential bugs is formal verification.

Formal verification, the cost of formal verification basically is going to plummet.

We're already seeing this with Aristotle, which is the large language model out of Harmonic.

That's the company that Vlad Tenev, co-founder of Robinhood, co-founded that basically is building theorem proving models.

You can use these theorem-proving models to do formal verification.

Ethereum already has a team internally that's doing formal verification on some of their cryptography.