On 2026-10-13 6:12 a.m., Laurent Pinchart wrote:
On Tue, Oct 13, 2026 at 01:59:33PM +0100, Lorenzo Stoakes (ARM) wrote:
On Tue, Oct 13, 2026 at 02:31:32PM +0200, Laurent Pinchart wrote:
On Mon, Oct 12, 2026 at 04:42:46PM +0100, Lorenzo Stoakes (ARM) wrote:
The tag's useful for push back on review,
That's the point I raised during the maintainer summit. It's actually
the *absence* of a tag that helps me push back on review. It allows me
to ask submitters if they forgot to add the tag, and in many cases (so
far) they simply don't reply.
A practical rule of thumb in the year 2026 is to assume LLM has been
used unless there is very compelling evidence to believe otherwise.
A voluntary tag only tells you that the submitter may have been aware of
the policy. It doesn't even prove that, because the tag could have been
added by the "agent" reading the documentation.
The decision to deprecate Assisted-by is right.
While the tag's useful for pushback, pushing back is the issue.
The asymmetry between submitter volume/reviewer bandwidth is the MOST
CRITICAL issue we face.
If reviewers are required to reply to all of these people, the project's
dead, AFAIC.
Ignoring is the only workable long-term solution.
How do you decide which patches to ignore ?
It is indeed the main issue.
netdev, for example, has an automated scoring system, partly based on
review to submission ratio - e.g. contributors who review other people's
patches are prioritized:
https://netdev-ctrl.bots.linux.dev/suie.html
https://lore.kernel.org/netdev/[email protected]/
A complementary thing is deploying automated patch review systems that
use LLMs, like sashiko. netdev and bpf have customized solutions and
they are quite helpful in terms of scaling the patch queue processing.
The issue we had before was a lack of a CLEAR and LOUD 'yes ignore these
people' from Linus.
We now have that :)
