The commit (3735384b193617a912fbf6d25844041ed288bf28) was indeed
completely erased from the master branch through a force push. That is
evidence that it was not perfectly logical to give it to Claude to
commit messages. Therefore, the defense that "Claude was only used for
the commit message" collapses under the weight of the actual response.
If that were true, the commit would have been reworded, not erased.
Developers are famously lazy in the efficient sense. If you're already
in your terminal, with git commit -m right there, why would you
interrupt your flow to involve a third-party AI service for a trivial
message?
The only logical answer is that author was already in the AI workflow
for the code itself, so getting a commit message was just an extra step
in that same session.
For the defense to be true, Yuan Fu would have had to:
1. Write the code by hand (fine)
2. Manually decide: "I will now use Claude exclusively for the commit
message"
3. Go to Claude, get the message, copy it, give instructions
4. Commit
But why? What was the purpose? There is no productivity gain. No quality
gain (it's a short message). No learning gain. It's an unnecessary
detour that only makes sense if the AI was already involved in producing
the code.
I love GNU and want it to remain free, which is exactly why I must point
out this flaw—not to attack, but to protect the project's integrity.
Jean Louis
---
via emacs-tangents mailing list
(https://lists.gnu.org/mailman/listinfo/emacs-tangents)
