Bill Gurley

And I think that's very doable.

I mean, there's some AI companies in the legal space.

I'm not involved in any of them, but they're studying all the...

intake information that you would put into a lawsuit, but they're also studying all the precedent and legal case history.

And the AI is going to do X. And if you've got a human in the loop, you're going to notice the failure points and then you're going to improve the model.

And those kind of things could lead to someone that has an early lead having an even bigger lead in the long term if there's constant improvement of the model.

There are people that have studied some of the stuff popping in China that hints at this.

And I should probably do more work on that and be more knowledgeable than I am.

I would also say we had an early shot on goal with character, which traded.

But there's two elements of the first wave of LLMs that I think make them not perfect for consumer.

One is voice being really good, which they're getting better at.

And the second one being memory.

And they're starting to get better at that.

I think they're doing it outside of the core LLM, but it doesn't matter.

And then they put it in the context window.

As those things improve, a lot of the negative feedback on the early character AI was that they didn't really get to know you.

And so the efficiency or the network effect and learning impact and switching costs weren't there.

It's not...

that difficult for me to imagine exactly what played out in her, which was an unbelievable movie for its time, by the way.

I think that will happen.