Post #647373
2026-03-10 01:54 UTC
The trick to block the Claude user on GitHub as a warning beacon is the gift that keeps on giving.
I’m not even that annoyed if people write code with AI, but letting the agent make commits for you feels like an admission that you haven’t reviewed its code.
A few weeks ago, a project I depend on broke some major behavior with a vibe coded commit. It’s been hard for me to trust repos with shamelessly LLM co-authored commits ever since.
(This post isn’t accusing Mermaid of anything, it’s just the latest repo I noticed it on)
Replies (1)
-
@bwebster@mastodon.social 2026-03-10 02:02
@kirb IMO this isn't necessarily a valid assumption, though I agree in the wild it probably has a high correlation to unreviewed commits. I use Claude to author my commits mostly because it writes much more detailed commit messages than I would ever be bothered to. 😅 I still always look through the changes first, it's just the final commit that I tell Claude to write. Hell I even tell it to author the commit sometimes for my own manual code changes!