And, and my, my question is, if not us, then who?

Yeah.

I, I, I think this is a, like this is a really big question in the, like in the context of the, the current wave of AI.

I think, um, there's two, there's sort of two forces that are countervailing.

One is, um, I believe in the importance of expertise, right? Uh, and that you get deeper and deeper into like, uh, sort of a modular breaking up of, of, you know, really hard challenging problems, right? If you can break it up into modules, and then you can get experts to get super, super deep, um, and really, really understand their individual module, they can improve their module, and then when you string the modules all together, then you get something great, right? This is a sort of traditional engineering approach to problem-solving, right? And it's really, really effective, as long as you can get the modules right, right? Um, you know, now, um, with AI, I think there's a major change in which of those modules are easy and which of those modules are hard.

And some modules that used to be hard are now easy, right? We talked about documentation, um-

Mm-hmm

... and I also think that it's coming for like optimization more generally, because optimization is, right, you can see how good the objective is, and that makes it a really easy target for recursive self-improvement.