Lennart Heim

So all of these things are now combined, but the cost can be kind of staggering, right?

And I think the key implication is here to achieve the best capabilities at the beginning, you spend a ton of time on appearance.

But again, AI moves forward, our exponentials move forward, everything gets cheaper over time.

You know, we do something and then we later learn how to do it cheaper.

Just like economies of scale, we build bigger models.

And I think that's the case for, like, computer science since forever, right?

And I think for AI, it's just, like, staggering because we have, like, these basically brute force approaches.

We've got a big model, we do it, and then later we get smarter over it and we have, like, these computer efficiency improvements.

And again, the odds are really fast, right?

It's roughly 3x per year.

So like basically the cost to achieve a given capability is like 3x cheaper at the end of the year because we just recruit.

Again, this is really hard to measure.

Do you do it with loss?

Do you do it on a specific benchmark?

It gets tricky, but the rough trend line is

And then DeepSeq is, I think, caught many by surprise because they just didn't think about it.

They just felt like, oh, it's always cost us a hundred million.

DeepSeq is perfectly on the trend line, basically, if you look at it, just like it sits directly where it's supposed to sit.

The exception might be test time compute.

This is actually a bigger increase in terms of compute efficiency in quantistar by improvement, which we've seen before.