How This Ex-Meta L8 Engineer Ships 40 PRs a Day with AI Agents | Kun Chen
Channel: Peter Yang
Uploaded:
Duration: PT3378S
Read a transcript preview of “How This Ex-Meta L8 Engineer Ships 40 PRs a Day with AI Agents | Kun Chen” from Peter Yang, uploaded on 2026-06-07.
Transcript preview
Language: English (original, YouTube automatic captions)
[0:00] If you review every single line of code,
[0:02] you become the bottleneck. So I don't
[0:04] reveal this first pass code from the
[0:06] agent. Eventually I got to a point where
[0:09] I find myself never catching anything
[0:11] the agents don't catch. I typically have
[0:13] like at least five different sessions
[0:15] actively running. On average, there's
[0:17] like 20 to 30 agents running. Most of
[0:19] the time, uh, it's like 20 to 40 kind of
[0:22] PRs every day. Our workflows and how our
[0:25] teams work were built at a time when we
[0:28] spend most of our time coding. But when
[0:30] you start to write like 10 times more
[0:32] PRs, we are not ready for that to really
[0:35] scale up how much we can get from the
[0:37] agents. We have to move ourselves out of
[0:40] the loop as much as possible.
[0:44] Hey everyone, today I'm really excited
[0:46] to welcome my friend uh, an L8
[0:48] engineer from Meta at Microsoft who's
[0:50] now a solo AI builder. is going to
[0:53] show us exactly how he builds products
[0:55] using agents. I've been asking him a lot
[0:57] of dumb questions about all this. So,
[0:58] we're really excited for him to show us
[1:00] live. So, welcome, sir.
[1:02] >> Thanks for having me here, Peter.
[1:04] >> All right. So, uh let's get right into
[1:05] it. Maybe you can start uh by kind of
[1:08] walking through at a high level how
[1:09] you're building products with agents.
[1:12] >> All right. Um that is my workflow. Um
[1:15] plan, code, and validates. Uh I don't
[1:17] think this is too different from uh what
[1:19] everybody does. Um, so I'll probably
[1:21] talk through the parts where I think I'm
[1:23] doing something unique. Um, so I think
[1:26] typically when we build something
[1:27] meaningful, w