Corey Noles

That's interesting.

Yeah, I think certainly from a UI perspective, that's true.

I want to double click on the idea of it being good for training data.

It's good for distribution and good for getting people into the ecosystem.

but it's your, your, you know, that's fair.

It's dangerous then to train on this because it's like, this isn't technically physics.

Yeah, exactly.

Like you want to be training on stuff that's like real to make it look even more real, I would think.

That's a really good call.

You could train it on the psychology and the types of prompts and types of videos that people want to make.

Yeah.

So that you can figure out like, what do we need?

How can we train the model to be better from like a prompt to video output?

And I think that's probably where the most value is going to come from is like basically analyzing...

These are the things that people want to make videos of.

These are the areas where they're having problems.

They're editing the prompt.

Like, you know, it's not coming out the way they wanted to.

So they change it multiple times.

You know, you can look at like the user behavior on the platform in order to figure out the areas that you need to fine tune and fix.