Ahmed El-Kishky

I think we announced it about a year ago, but the biggest leap was just the introduction of reasoning models.

If you remember when ChatGPT was released, I guess almost three years ago, ChatGPT would ask you a question and immediately it would just give you an answer.

It would just always give an answer.

Sometimes the answer would be correct.

Sometimes it wouldn't be.

It didn't matter how hard the problem was.

It would just always immediately give you an answer.

Yeah.

So back then, people had these like approaches.

They called it like chain of thought prompting.

They'd be like, show your steps, like think through step by step.

They would just ask the model to.

And people noticed that when you just asked it to like, you know, show your work, the benchmarks would just improve a little bit.

It'd be a tiny bit smarter.

Yeah.

So we worked on that.

We were like, OK, what's happening here?

And the idea is that like when you have a difficult problem, no human just immediately like gives an answer.

If I asked you to multiply like a four digit number by a five digit number, you wouldn't like give me an answer instantly.

You would like, you know, get your piece of paper and pencil.