So, uh, Roy, what's your sense on that? What, what does, uh, AI, uh, native language look like? And also, uh, what, what is your, uh, continuous learning? And, uh, Michael has mentioned, uh, uh, the place into this big picture.

Yeah.

I would say what AI native actually means has been also changing really rapidly.

If you asked me like six months ago, it would probably be, you know, organizations who are probably using AI for like knowledge work and increased productivity, and that has even dramatically shifted in the last six months.

Right now, if the, if you have like a, uh, if your teams don't have maybe like 100 or so agents that are doing a lot of things in the background while you are focused on like the more higher creativity tasks, then I would say you're probably not necessarily fully AI native, right? So the bar of what AI native means has radically shifted.

Uh, in terms of like the organizational structure, um, I think, uh, one of the things which has changed is, um, the traditional barriers for specific roles is in some ways eroding.

Because, uh, if you're a product manager, you can suddenly actually develop a prototype very easily, so you don't necessarily need to be an engineer.

Then the, uh, uh, would you like to talk a, a little bit on the auto scientist that you are working on now?

Mm-hmm

... should be adapted to the complexity of the task.

If you're writing a simple paragraph or editing a copy, um, um, uh, for, I don't know, social media, then, uh, you don't necessarily need to spend huge amounts of compute in order to accomplish that.

They're obviously, like, much more complex.

Maybe if you're doing a medical diagnosis, then it may be worth it to actually spend massive amounts of compute.

So how can you do the proportional allocation of compute depending on task complexity is one example of, uh, adaptation, for example.

Um, the other thing that's worth noting is that we are going through a phase where we see lot of potential for AI, and that comes, uh, with a lot of demand for compute, but also a lot of demand for energy, uh, that is needed in order to, like, um, drive that compute.

Maybe what's a mental hurdle that you overcame that maybe you could shed some light on maybe how they could also overcome that?