AI-assisted review

[Josh Law]

On May 25, 2026 3:27:18 PM GMT+01:00, Tom Rini <trini at konsulko.com> wrote:
>On Mon, May 25, 2026 at 03:25:19PM +0100, Josh Law wrote:
>> On May 25, 2026 3:21:59 PM GMT+01:00, Tom Rini <trini at konsulko.com>
>wrote:
>> >On Mon, May 25, 2026 at 03:11:49PM +0100, Josh Law wrote:
>> >> On May 25, 2026 3:03:28 PM GMT+01:00, Simon Glass <sjg at chromium.org>
>> >wrote:
>> >> >Hi Tom,
>> >> >
>> >> >On Mon, 18 May 2026 at 09:58, Tom Rini <trini at konsulko.com> wrote:
>> >> >>
>> >> >> On Mon, May 18, 2026 at 10:55:40AM +0200, Michal Simek wrote:
>> >> >> >
>> >> >> >
>> >> >> > On 5/16/26 00:07, Tom Rini wrote:
>> >> >> > > On Fri, May 15, 2026 at 03:03:21PM -0600, Simon Glass wrote:
>> >> >> > >
>> >> >> > > > Hi,
>> >> >> > > >
>> >> >> > > > There was a query on the call this week about whether I am
>> >doing
>> >> >> > > > AI-assisted code review. As I said on the call: yes. Here is
>a
>> >> >brief
>> >> >> > > > description of how it works.
>> >> >> > > >
>> >> >> > > > It is built into Patman (on the Concept tree) with a new
>> >'patman
>> >> >> > > > review' command. You give it the series name / number, or
>> >perhaps
>> >> >a
>> >> >> > > > patch name/number and it applies the patches to a new
>branch,
>> >does
>> >> >a
>> >> >> > > > review then adds its comments to its database.
>> >> >> > > >
>> >> >> > > > A '-d' flag can be used to create draft emails in Gmail
>(sorry,
>> >it
>> >> >> > > > doesn't support other email programs yet). You then check
>and
>> >> >update
>> >> >> > > > the emails and send them (or delete them). I am not an
>expert
>> >in
>> >> >> > > > handling the 'user voice' part of AI, but have made an
>attempt
>> >to
>> >> >make
>> >> >> > > > it follow any provided configuration, as well as to scan
>recent
>> >> >> > > > reviews to actually create to create a voice.
>> >> >> > > >
>> >> >> > > > Obviously this is very rudimentary and could be expanded
>> >> >considerably.
>> >> >> > > > But the mere fact that it creates draft emails is a win for
>me,
>> >> >even
>> >> >> > > > if I ultimately delete or rewrite most of the comments. I
>can
>> >> >imagine
>> >> >> > > > 10 different ways to improve it to be more useful.
>> >> >> > > >
>> >> >> > > > I wrote a blog post about it if you want more details, or
>you
>> >can
>> >> >ask me here.
>> >> >> > > >
>> >> >> > > > I am very interested in hearing how others are using these
>new
>> >> >tools
>> >> >> > > > for code review.
>> >> >> > >
>> >> >> > > And the big thing for now is that since we as a project do not
>> >yet
>> >> >have
>> >> >> > > an AI policy aside from "please don't". One of the points I
>was
>> >> >making
>> >> >> > > on the call is that there's a difference in value between
>"Human
>> >> >> > > reviewed it, looks fine" and "Human spent some tokens, agent
>> >didn't
>> >> >see
>> >> >> > > any problems".
>> >> >> > >
>> >> >> > > And I know several other people have been doing at least first
>> >pass
>> >> >> > > reviews with various agent-tools, it's just no one else has
>been
>> >> >posting
>> >> >> > > reviews at your scale. And lessons learned from other projects
>is
>> >> >that
>> >> >> > > the prompts are more important than whatever wrapper around
>the
>> >> >agent
>> >> >> > > one is using.
>> >> >> > >
>> >> >> >
>> >> >> > Don't think scale is the problem. Tool and integration is
>another
>> >> >topic.
>> >> >>
>> >> >> Simon posted approximately 100 reviews in about 24 hours. That
>scale
>> >is
>> >> >> a problem, when most of them are just reviewed-by tags, from
>someone
>> >> >> that has a history of doing human reviews. Reputation is a factor
>> >here
>> >> >> I'm trying to figure out how best to articulate.
>> >> >>
>> >> >> I have thoughts on the rest that I want to get back to later,
>thanks.
>> >> >
>> >> >I should point out that I tend to do reviews locally bit by bit and
>> >> >then recheck and send out in batches later, particularly when I need
>> >> >to dig into the code and check things. I suspect a lot of the
>> >> >'reviewed-by' ones are on revised series where I already reviewed
>v1,
>> >> >etc. For better or worse, patman tends to have something to say on
>> >> >most patches (too picky for my style so I often delete comments).
>> >> >
>> >> >Re the AI policy, I suggest adding it in the project docs (even if
>it
>> >> >is very brief), rather than referencing a URL from another project.
>> >> >
>> >> >Regards,
>> >> >Simon
>> >> >
>> >> 
>> >> Hey guys, sorry for the unexpected email but I have a question
>> >> 
>> >> How would you know the reviewed by tag wasn't just made by a AI
>> >> 
>> >> Would like: "Here is the tag from soandso AI"
>> >> 
>> >> Or would it be like
>> >> 
>> >> "AI reviewed this and it looks fine"
>> >> 
>> >> Apologies for the unexpectedness of the email :)
>> >
>> >That's one of my concerns, yes. And I've been a bit shocked that other,
>> >bigger, projects that do allow for AI review haven't come up with
>> >something already.
>> >
>> >
>> 
>> Personally Tom, projects like Linux take it slightly different
>> 
>> maintainers usually say:
>> "AI asked a question" then link the sashiko (AI review tool) link, with
>> all of AIs comments
>> 
>> The general gist, is if there is any comments, you either fix it, or you
>> say why it isn't a bug
>> 
>> I have mixed opinions on sashiko tbh.
>> 
>> Also, maintainers tend to use their **OWN** review tools.
>> 
>> But most projects do generally ban AI anything.
>> 
>> Idk what we could do for u-boot when it comes to a sashiko like
>solution.
>> 
>> Maybe we could talk about it?
>
>Proposing a policy for the project about AI (and indeed not just linking
>to other projects which have what I find to be helpful references) is on
>the TODO list for the project leadership committee. We've just been
>busily handling other issues for a while now, unfortunately.
>
>

I could put something into suggestion:


So, for patches: you could do assisted bys, and it depends on the
maintainer, etc etc like Linux,


And for reviews: that's actually hard to think about, because some
maintainers love using AI review tools because it's useful, others
don't particularly like tools like that, maybe a opt-in "reviewman"?

But tbh that's up for discussion.

Thanks!