Hi,
On 7/21/26 3:43 PM, Shawn Rutledge via Development wrote:
>> As a reviewer I try to find out quickly whether AI was used as
a tool
>> to enhance productivity, while the developer owns the code.
>
> I think anything not a small patch - medium and large patches,
should be
> clearly identified in the commit message, that it was developed in
> conjunction with a machine or somesuch.
I'm not convinced this would be useful. Such a disclaimer would just
serve as excuse should the developer not fully understand / own the
generated code, which is the situation we should avoid. I'm also not
interested to read in commit messages which IDE, text editor,
operating
system was involved in creating the patch. ;)
That's not an apt comparison IMO.
If you used some code generator (from some IDL, protocol
specification, introspection or whatever), you would usually say that
it's generated code. Sometimes that's considered a build artifact to
be regenerated every time; sometimes we check it in with a comment
about how it was generated, because we don't expect that it needs to
be regenerated very often.
It seems to me the Linux community is setting the standard here:
https://docs.kernel.org/process/coding-assistants.html
They say the tagline should be Assisted-by: ...
Whereas Claude is doing this by default:
Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
That does look too much like an ad for my taste, and co-author is a
bit stronger word than assistant; but, maybe we'll appreciate the
model version when we look back later: ah yeah that 4.8 version was
not so smart, it made this typical kind of mistake that we know how to
correct now, or something like that.
FWIW that's the KDE position. Assisted-by and Co-authored-by act more
like an ad here. You typically don't put Developed-with: Visual Studio
Code, etc.
Regards,
Vlad
I think it's a matter of honesty to tell when an AI agent has done a
good part of the work, just as you would add Done-with: when a
colleague or collaborator has done a substantial portion of it. It's
not a matter of abdicating responsibility, it's "show your work" the
way every academic expects. Give credit where credit is due, and also
show the line of reasoning that led to the code being the way it is.
IMO the reason we write long commit messages is so that "git blame"
can answer questions later on: why is this code the way it is? What
frame of mind did the person who wrote this have? What bug was he or
she trying to fix? AIs often explain their thinking more thoroughly,
in commit messages and comments. OTOH so far they do not return the
favor of telling which human sources inspired any particular output.
It would be nice if they did, but they apparently don't know how to
remember the sources that they learned from.
Confidential
--
Development mailing list
[email protected]
https://lists.qt-project.org/listinfo/development
