Karl Yeh

You got like chain of verification.

He has an adversarial prompting straight edge use case learning.

Like some like even like, you know, I think some of us have done the meta prompting where you're asking it to prompt.

But but there's a lot of ways that you can solve a problem.

But I think what Rachel instilled in all of us, because we're all in the same community way back when, is really the systems thinking part of it.

Like, how do you break this down?

And then how do you apply some of these models that isn't a straightforward... Now, ideally, the tool will get better to breaking it down for you.

And because...

memory I think it'll you know understand the context a little bit better instead of starting from scratch every time but some of these techniques still super super useful to have that right it's like and then also to recognize hey this hallucination or or this thing is

is not right, and maybe a simple, hey, just if you don't know, you don't know.

Like, don't make it up.

If you don't know, you don't know.

That's cool.

That kind of stuff.

So I don't know.

It's a tough, I don't know how to, like, there's no best practice way how to apply this in a space that's moving so fast where you need to not just know the capabilities, but also using it on a daily basis.

But at the same time, too, those capabilities could change and get gobbled up.

And you're like, well, I have to relearn that capability because it's that combo that I used to do is actually more efficient.

But I didn't know that there was another piece that made it even better.

But I just wasn't paying attention.