I am asking my team to flag git commits with a lot of LLM/Agent use with something like:
[ai]: rewrote the documentation ...
This is helps us to put another set of "glasses" as we later review the code.
[ai]: rewrote the documentation ...
This is helps us to put another set of "glasses" as we later review the code.
If you use AI as tab-complete but it's what you would've done anyway, should you flag it? I don't know, plenty to think about when it comes to what the right amount of disclosure is.
I certainly wish that with our company, people could flag (particularly) large commits as coming from a tool rather than a person, but I guess the idea is that the person is still responsible for whatever the tool generates.
The problem is that it's incredibly enticing for over-worked engineers to have AI do large (ie. diffs) but boring tasks that they'd typically get very little recognition for (eg. ESLint migrations).