Connect your Claude, Codex or OpenClaw for full transcripts on millions of podcasts. Two week free trial, cancel anytime.
Start free trial →Front-End Fire · Hands-On Engineering Podcasts · October 2026
The hosts were discussing how code review works when agents write most of the code, including a proposal that reviewers be given the AI session behind a change. Niedringhaus says her team instead commits plan files where the agent records what it actually did, and keeps PR descriptions short.
…just want to look at the code or smoke test it, whatever, then you can just read the PR description that is short and succinct and something that is small. Yeah, but it is there for posterity. It's there for future agent sessions. Like the agents can look at it all they want, learn things from it.
…you. I mean, if you solve this, let us know in the YouTube comments or on our Discord because we're always curious what other people are doing. Because I don't think anybody really knows. And the best we can do now is just talk about it and share stories, share what's working for us, share what's
not. Yeah, but don't share your session with me. I do not want to read that. I'm not going to read that. It's just the honest truth. I don't read all of my clawed sessions myself, so I'm not going to read somebody else's.
The AI slop is getting out of control. This is the other thing that's bad in our modern era is now you'll get these like AI slop files that somebody else in your company will just send you. And you know, it was just, you just had Claude just mass create this. And then you're just, when you're, when…
Yeah. I will say, in defense of the agents, at least this week in particular, there's been some big updates to the Claude desktop app where if I've started on one session and I'm working on something and I need to continue a new session because it's a fresh, you know, it's a related issue or it's…
Yeah, I found that both Codex will start doing this now too, where it'll spawn sub-agents. And it's almost kind of mesmerizing to watch because they'll say like, oh, well, I can help you, but I'm waiting for this sub-agent to report back. It's off doing some research and it's wild.