Scale AI Development With Engineering and Agent Control Planes?

Uh, heap engineering.

Uh, yeah.

Um, I think it's an interesting, um, topic.

Uh, it's interesting to see, uh, what's the discussions that's happening on Twitter and there's some meta commentary here about like,

you know,

how do we get our news about what's happening on AI? Um, and I, I wanna say that like Twitter's not, you know, the one place where everything is happening.

There's a lot of quiet.

conversations going around so I agree with you there's a lot of hot takes of like you know loop engineering is dumb like absolutely nobody running a serious company is doing anything like loop engineering nobody's running their agents overnight and then you have these debates and we I saw a lot of this debate like two months ago and then we are seeing tweets now how like somebody's just like merging thousands of PRs per week or something like that.

And so the discussion is just all over the place.

And I think there's a lot of people who are very opinionated about like, no, I cannot imagine how anybody could possibly be doing this.

Therefore, everybody else is full of BS.

And then somebody who's quiet comes out with these amazing results about what they're getting.

the agents to do, and we kind of have to take that seriously.

So I think loop engineering is one of those things that I've been interested in because I don't want to think that much.

I mean, I do want to think I do want to use my critical thinking, but I don't want to be using my critical thinking in the same way that I was using it as a staff software engineer.

So I can give you an example.

Recently, we turned this internal project into what's going to be our main product called Firehose.

And that is our agent manager.

But initially, I just made it as a regular React app.

And then I was like, man, I should have just made this into an expo app because then I could just make it into a universal app.

And so right around when Fable came out, I was like, OK, I want to see if I can get Fable to just one shot this whole thing because I heard it's amazing.

And it turns out that it did not work out.

And so as I started to think about this problem, Yeah.

You start getting into the details, right? Our product was already very, very complex.

There was a lot of stuff going on.