Re: [PATCH] docs: remove Assisted-by references

[lyude]

On Tue, 2026-10-13 at 13:45 -0700, Linus Torvalds wrote:
> 
> I don't understand this whole idiotic concept of "don't use AI for
> X".
> 
> Here's a new rule just for you: don't use compilers for any of your
> kernel submissions.
> 
> Because if you can't show that you really understand what the machine
> is actually doing by writing your code in assembler, then I don't
> want
> to see your code contribution. You clearly aren't competent if you
> can't write it that way.

I don't think this is a honest comparison when there has been a very
clear shift in the kind of janitor patches I have received over the
last 3 years. It used to be only occasional I got patches from people
that they managed to write and build, but didn't understand. It is
incredibly frequent now, and there is a very noticeable trend of what
kind of tools are used for generating those patches. At the dawn of
programming languages, sure, people could keep throwing things at the
wall without understanding what they were doing. They could do it
somewhat faster then before. But it still took time, it was not
instant.

I've spent most of my career trying to push for new tooling, I pushed
pretty hard in regards to pushing rust in the kernel. I've pushed for
gitlab, I've pushed for new build systems, I literally believe that
change is a part of what makes a developer good at their job. If you
can't accept change happening anywhere around you, you blind yourself
to new solutions that could make your work better. I even have started
using LLM review bots like Sashiko, despite the fact that personally I
wish these things never existed.

And even I am not going to go out and try to say "it's just a new
tool", because that's incredibly short sighted and just false.
Usefulness doesn't change the fact we do not have the same social
contracts we did before LLMs were around. And when a "new tool" changes
the de-facto social contracts that a project has had, that seems quite
a bit different then introducing a static analysis tool, a new API, a
new programming language, etc. into the kernel. Rust helps attract new
contributors, but it's never fundamentally changed the way that I've
handled reviewing people's work or handle working with new
contributors. It's perhaps saved me some time, but it's never
fundamentally changed the expectations I had when a new contributor
appeared. LLMs absolutely have changed that.

This is the distinction between LLMs and other tooling, and it is a
large enough one that at the very least it should be addressed with
stronger arguments and a higher level of respect then "You just hate
new tools don't you?".

Frankly - that's part of why you don't understand "Don't use AI for X".
You keep trying to look at this like someone introduced a simple new
tool, and now people are just being luddites. If that's the only way
you can approach these discussions then yes, you're never going to
actually understand what's going on.