Andy Halliday

You know how Jensen, he loves to make an announcement, but he's already got this working behind the scenes.

Well, there's a list of about 15 or 20 major companies around the world that are already using the Apollo model lines in their sort of

cutting edge development of these things.

So you've got Prometheus over here at Amazon working on the world of physical modeling.

And you've got now NVIDIA's Apollo family of models, which is all about physics and physical space modeling.

So that's, this might, we've thought, you know,

2025 is really the year of agents, and I think that's true.

We've seen a lot of advances, and we're wrapping up the year with a wide range of agentic capabilities.

At the same time, the next wave is clearly going to be, and maybe 2026 is all about world models and physics.

And so that gave me encouragement about NVIDIA's

forthcoming highly newsworthy announcement after the close of markets today of their third quarter results uh and and the entire s p 500 the major weight of which is nvidia's stock alone uh it's waiting to see whether there's you know going to be a conclusion by investors there's a bubble and you know things are actually retarding and slowing down or whether

The signals are inside NVIDIA's performance in the third quarter that say, no, no, no.

This, along with many other forces, are really pushing us dramatically forward in the world of AI.

Let me mention one thing about, you know, the knowledge cutoff dates.

You notice there that the knowledge cutoff date is the same for for Gemini 3 as it was for 2.5 Pro.

January of 2025, almost a year ago now.

And I think that the point I want to make is that it's no longer so important what the pre-training data cutoff was because now the advances that we're getting in the models are less about scaling the data ingestion

and less about the recency of the data in order to provide up-to-date information.

Why?

Because all of the advances that we're getting in reasoning don't have much to do with the data, which is just the basic building blocks of learning the language.