Illia Polosukhin

on itself, I think the challenge is, like, if you think of, like, what you want from this model, like, you want a lot of background processing, right?

You want it to, you know, read all your news and be able to, like, you just cannot expect all this to happen on your device, even assuming the model is great, right?

And so that's why we kind of like, because we were initially looked at local models and kind of small models, edge intelligence, we called it.

And yeah, it wasn't clear, like, again, that's why I look at the small models and kind of their progress.

But it wasn't clear that that's going to be able to satisfy the product requirements, right?

And especially as we think of businesses wanting to use this, there's just background workflows and analysis of data and all those things you want.

But it's still really important, I think, to continue making progress on smaller models because it just makes things cheaper, if nothing else.

Yeah.

It's always good to talk to a smart person, even if the person is AI.

At near protocol on X at the old black dragon, uh, for me personally, uh, and then near the dork is the website.

Uh, and there's blog there with lots of posts.

And then I actually just started a sub stack, no black dragon.com for more, I would say like longer term things that are maybe not like not as relevant to what we're working on right now, but I think pretty critical in this AI space.