On Tue, 17 Nov 2026, Philippe Mathieu-Daudé wrote:
On 17/11/26 22:42, BALATON Zoltan wrote:
The problem is not people using AI but people using AI who otherwise have
no idea what they are doing and send the slop unreviewed and untested (or
submit tickets that way)
I'd rather blame AI-induced automation bias which also happens to
"people who have idea of what they are doing".
That's a second level of problem we need to deal with once we manage to
exclude the non-sensical slop or people who submit tested AI patches but
then don't reply to review because they don't understand what the patch
does or review asks. Then it may become a problem that it's easy with AI
to make large changes touching a lot of areas which is much harder to
review. But if these are submitted by people who know what they are doing
we can at least expect them to read and follow the policy and review, test
and understand the changes before submitting so it will not be stuff that
nobody understands or can correct according to review. This still means
more work for reviewers but at least they don't have to fight the slop
first to find the valuable changes which should be done by the submitter
in the first place. If such large changes that become possible with AI
used by people who know what they are doing helps improving the project
it's probably worth the increased review work. So I still think we should
try to filter submissions that are sent by people who don't know what the
patch does and cannot answer review because then review work is wasted
effort.
Regards,
BALATON Zoltan