But Sergey, I'll let you go into each of those.

As, uh, Pen said, yeah, indeed, we had to do, like, a multilayered approach because there's two areas.

One is the player safety, so that they don't experience something they didn't, you know, sign up for.

But there's also talent safety, right? Making sure talent is confident we, we've got their back and their, let's say, avatar, their representation in our game is behaving the way.

And what's super important is so-called red team destructive testing.

It's when we send, you know, tens of thousands of variations of possible scenarios that would trigger the model, and we test against that to ensure model behaves the way we intend it to.

So all of that actually takes quite a big chunk of time for the development, so that's what is a characteristic of Gen AI game development.

So that person that just got their Choco Taco, made it back in, they're getting ready to open up their laptop and start using generative AI, what's one practical technique that they can use or that you've used and found to be valuable today?