Karl Yeh

I'd look through it.

But just do you see how fast that was?

So what I asked it to do was I wanted to take the skill, just the skill that I built from Claude.

I would take the discovery session analyzer skill.

And what that skill does is it goes through the summary and the transcript of a one and a half hour session we have with a department with one of our clients and

scrapes that reviews that analyzes it.

I have a format that I want to put it in.

So it puts it in that format and then it goes off to search, um,

In addition to some other things that I tell it to do and come back with non-AI solutions, potential AI solutions, which I still have to review, roadmap and everything.

And that's usually, again, it takes about two, three, four, five minutes for, because I usually use ChatGPT Extended to do that review.

I use Cloud Opus 4.6 to do that review just to give me two.

I did it in Codex Spark and I was like, holy smokes, that that was you counted.

That was under 10 seconds, I think.

Yeah.

So.

Yeah.

So technically, so Greg, anything you do with cloud code, try it with Codex.

Mind you, Codex is super fine-tuned to actually code, so why would you do it?

But the key thing here is they're testing Cerebrus on like a very, very niche coding, but they want to make sure the infrastructure is right.

But could you imagine this in ChatGPT?