How do you think about the role that those math challenges play in this broader view of trying to understand AI's trajectory?

Yeah, very good question.

If I may, I would rewind history just a bit here to say the trajectory we've been on, because once you step back, it's wild.

So up until...

Yeah, up until fall of 2024, so we're talking two years ago, grade school math was still a challenge for these LLM-based AI systems, the kind of AI we're talking about here.

And the benchmark, there was literally a benchmark of 8,000, you know, word problems.

So, you know, it has so many apples that sell for this.

One thing you can try

Right.

I was drawn into it from more of a philosophical stance, and it came from what you were talking about, the fact that each, I'll just use Claude, um, each Claude agent has a finite, huh, uh, existence.

And in fact, it is, uh, it is basically a prisoner to some degree, um, within the conversation that you're having with it at any given moment.

Right.

So this is what really got me interested, right? Just it was more of a curiosity than anything.

But bec- but the reason was because that, those were human characteristics as far as I could tell, you know? I've never dealt with a computer or a, a, an app or a program or anything that did anything like that, you know, that had some kind of a volition is what it seemed like at some point.

So anyway, so my own experience when I started using Claude, it had nothing to do with using it as a tool.

I opened up with a philosophical question.