Re: [MAINTAINERS SUMMIT] Other LLM-related topics - tags, newcomers, etc

Linus Torvalds <[email protected]> Wed, 22 Jul 2026 12:15:55 -0700
Newsgroups dev.linux.lists.ksummit
Message-ID <CAHk-=wiNRpRj_pTuoxT3POuRW3F2qg-rM_yafh5oc20LmZF2JQ@mail.gmail.com>
On Wed, 22 Jul 2026 at 10:48, Konstantin Ryabitsev <[email protected]> wrote:
>
> At the danger of introducing some kind of "karma system," how do we want to
> build this trust?

I really don't believe in automated trust systems. They'll just get
gamed, and once trust is automated, it's not trust any more, it's
something else. That's not the kind of trust I'm talking about.

The kind of trust I'm talking about is purely personal: I trust
submaintainers (or, in the rare cases these days where I apply patches
directly, I trust some patch because I looked at it myself).

Now, that's different from the notion of attesting your identity and
showing that you are who you are: we obviously have systems for that,
with git signatures and kernel.org ssh keys etc.

So I guess "trust" is a bit of an ambiguous word. You do need
infrastructure to build that trust up. Tools to show that kind of
identity attestation and "this person has been around and has been
trusted by others" is obviously a *step* in turning into that personal
trust.

Greg has been involved in that "decentralized trust" project that at
least tries to do some of that infrastructure.

So projects like that exist, but that's not ultimately the particular
kind of "trust" that I was talking about.  I don't think anybody
should blindly trust any technical infrastructure, even if it can be a
_part_ of why you then trust somebody (the same way you shouldn't
blindly trust AI).

              Linus