The AI PR that made the errors go away with a try/catchI keep seeing the same shape of agent PR. Something throws. The agent "fixes" it by wrapping the call in try/catch, logging a vague message, and returning a default. CI goes green. The PR description Oct 12 · 1 min readrivendesk.hashnode.dev
The AI PR with no rollback storyThere's a class of AI-written PR that sails through review for a dumb reason: nothing looks wrong. Diff is medium. Tests are green. Description sounds careful. Nobody can point at a bug. And still, atOct 10 · 2 min readrivendesk.hashnode.dev
The AI PR that fixed the test instead of the bugHere's a pattern I keep running into with agent-written PRs, and it's sneaky because everything looks fine. The task is something like "fix the date formatting bug on the invoice page." The agent make6AOct 7 · 3 min readrivendesk.hashnode.dev
Green CI isn't a merge decision: 8 times I stop an AI-written PRMost of my review time on agent PRs used to go into reading the summary, glancing at the diff, seeing green checks, and approving. That's fine until the one PR that quietly bumps a lockfile, renames a2IOct 5 · 3 min readrivendesk.hashnode.dev
Free one-pager: STOP conditions for AI agent PRsAI-assisted pull requests can be smaller and safer—but only if “looks plausible” is not treated as a review. STOP is a compact pause point for AI-authored PRs. It is not anti-AI; it is a reminder to sSep 11 · 2 min readrivendesk.hashnode.dev