Grant Harvey

Yeah, like custom styles, like Anthropic has that, right?

Like, yeah.

Or you can give it your custom style and say, I want you to write in this style, and you can give it as many instructions as you want.

Yeah, that's fair.

Okay, got a random one for you before we go.

Why do the models love em dashes so much and colons and titles?

Why can't they write a complete sentence?

I feel like they hate complete sentences.

Yeah, because then it gives you the human indication, like, okay, this is slop.

Yeah, it ensures human in the loop because you're editing out all the em dashes.

Oh my gosh.

Is it possible, do you think, with the current architecture to train for diversity?

Because right now, my understanding is that it's sort of like the models lean into the middle, right?

It leans to what's most in distribution, like to your earlier point about lack of diversity in the train data.

Can you ever actually steer it towards diversity and having diversity be a key part of it in terms of like,

So your sentences always start a different way.

It's always the creativity part of it, I guess I'm wondering.

Maybe Serge is involved.

Maybe it's his job.

I bet you're right.