Ahmed El-Kishky

really cool and they're incredibly excited by it and they want to see how far they could push this.

And I think it really, it's along our goals as a company.

We want to sort of see how far we can push the limits of autonomous, just agentic work.

And that involves both thinking, reasoning, but also engaging with, for example, a computer or an environment.

We think that's sort of the next frontier.

Competitive programming, our models are now among the best.

For IMO, we've shown that, hey, we can play at the same level as some of the best mathematicians in high school, and these are some really tough competitions.

The next frontier we're thinking about is what if these models took longer than hours?

What if they took days, weeks, months to even solve a problem?

Some of the hardest problems in the world I don't think are going to be solved in hours.

I agree.

Yeah.

If you want to further sort of human knowledge...

Almost like a PhD student would spend four or five years working on this very narrow subject with the goal of just like, you know, making a tiny increment into, you know, the sphere of human knowledge.

We're hoping our models sort of do the same thing.

Uh, we want to give them difficult problems and we want them to maybe like you asked a question and then you leave, you, uh, come back like in a day or two, or maybe like you get an email.

That's sort of where I personally feel the world is going to.

Um, you're going to have these be a amazingly powerful tool, um, that can, you know, solve way more difficult problems.

Um, one thing that I'm really excited about is like AI for science.

Um,