[PATCH] coding-assistants: simplify attribution

[PATCH] coding-assistants: simplify attribution

[Jonathan Corbet]

As suggested by Christian, I have applied his patch to a branch and
simplified the changelog, see below.

I'm nervous about applying it still because I don't think it's my role
to set this sort of policy; I really want to feel that there is a
consensus behind it.  So there's some additional CCs here because I
would like to hear from some folks:

- This policy document originally came out of the TAB; is the TAB (on
  CC) OK with this change?

- At least one maintainer (on CC) have said that they currently strip
  Assisted-by tags out of patches before applying them, a practice that
  would seem to undermine the whole policy.  Would this change be
  sufficient to end that practice?

Thanks,

jon

From c31aa376a355e3b84b9e87ae1dbe19b394d4c82c Mon Sep 17 00:00:00 2001
From: Christian Brauner <brauner@kernel.org>
Date: Wed, 1 Jul 2026 17:54:48 +0200
Subject: [PATCH] coding-assistants: simplify attribution

The requirement to identify specific models used in the Assisted-by tag
provides free advertising to proprietary software companies while adding
little or no useful information.  Change the requirement to simply:

  Assisted-by: LLM

to capture the fact that an LLM was used without tracking which one.

Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
Reviewed-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Jori Koolstra <jkoolstra@xs4all.nl>
Message-ID: <20260701-work-coding-assistants-v1-1-a20a94d1d606@kernel.org>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
---
 Documentation/process/coding-assistants.rst | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/Documentation/process/coding-assistants.rst b/Documentation/process/coding-assistants.rst
index 899f4459c52d..fe34f3e7e828 100644
--- a/Documentation/process/coding-assistants.rst
+++ b/Documentation/process/coding-assistants.rst
@@ -43,12 +43,8 @@ When AI tools contribute to kernel development, proper attribution
 helps track the evolving role of AI in the development process.
 Contributions should include an Assisted-by tag in the following format::
 
-  Assisted-by: AGENT_NAME:MODEL_VERSION [TOOL1] [TOOL2]
+  Assisted-by: LLM [TOOL1] [TOOL2]
 
-Where:
-
-* ``AGENT_NAME`` is the name of the AI tool or framework
-* ``MODEL_VERSION`` is the specific model version used
 * ``[TOOL1] [TOOL2]`` are optional specialized analysis tools used
   (e.g., coccinelle, sparse, smatch, clang-tidy)
 
@@ -56,4 +52,4 @@ Basic development tools (git, gcc, make, editors) should not be listed.
 
 Example::
 
-  Assisted-by: Claude:claude-3-opus coccinelle sparse
+  Assisted-by: LLM coccinelle sparse
-- 
2.55.0

Re: [PATCH] coding-assistants: simplify attribution

[Konstantin Ryabitsev]

On Wed, Jul 22, 2026 at 02:34:40PM -0600, Jonathan Corbet wrote:
> -  Assisted-by: Claude:claude-3-opus coccinelle sparse
> +  Assisted-by: LLM coccinelle sparse

I've been playing with this a bit, and it's pretty easy to teach the agent to
add a few tag-like notes as to how exactly it was used, e.g.:

    https://git.kernel.org/pub/scm/utils/b4/b4.git/commit/?id=018b87444d32eb2ffee7cb25f0785e54aa35f7bf

This has:

| Assisted-by: LLM [design, codegen, tests]

It tells us a bit more about *how* the LLM was used, without introducing the
advertisement or pinning exact models.

This format should allow for multiple tool usage:

| Assisted-by: LLM [triage, tests], coccinelle

I don't suggest this is included in this version, but it *is* fairly
straightforward to add to AGENTS.md and tells reviewers just how exactly the
LLM was used.

-K

Re: [PATCH] coding-assistants: simplify attribution

[Mauro Carvalho Chehab]

On Wed, 22 Jul 2026 16:48:03 -0400
Konstantin Ryabitsev <mricon@kernel.org> wrote:

> On Wed, Jul 22, 2026 at 02:34:40PM -0600, Jonathan Corbet wrote:
> > -  Assisted-by: Claude:claude-3-opus coccinelle sparse
> > +  Assisted-by: LLM coccinelle sparse  
> 
> I've been playing with this a bit, and it's pretty easy to teach the agent to
> add a few tag-like notes as to how exactly it was used, e.g.:
> 
>     https://git.kernel.org/pub/scm/utils/b4/b4.git/commit/?id=018b87444d32eb2ffee7cb25f0785e54aa35f7bf
> 
> This has:
> 
> | Assisted-by: LLM [design, codegen, tests]
> 
> It tells us a bit more about *how* the LLM was used, without introducing the
> advertisement or pinning exact models.
> 
> This format should allow for multiple tool usage:
> 
> | Assisted-by: LLM [triage, tests], coccinelle
> 
> I don't suggest this is included in this version, but it *is* fairly
> straightforward to add to AGENTS.md and tells reviewers just how exactly the
> LLM was used.

I'm fine with either way.

Reviewed-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>

Regards,
Mauro

Re: [PATCH] coding-assistants: simplify attribution

[Lorenzo Stoakes (ARM)]

On Wed, Jul 22, 2026 at 04:48:03PM -0400, Konstantin Ryabitsev wrote:
> On Wed, Jul 22, 2026 at 02:34:40PM -0600, Jonathan Corbet wrote:
> > -  Assisted-by: Claude:claude-3-opus coccinelle sparse
> > +  Assisted-by: LLM coccinelle sparse
>
> I've been playing with this a bit, and it's pretty easy to teach the agent to
> add a few tag-like notes as to how exactly it was used, e.g.:
>
>     https://git.kernel.org/pub/scm/utils/b4/b4.git/commit/?id=018b87444d32eb2ffee7cb25f0785e54aa35f7bf
>
> This has:
>
> | Assisted-by: LLM [design, codegen, tests]
>
> It tells us a bit more about *how* the LLM was used, without introducing the
> advertisement or pinning exact models.
>
> This format should allow for multiple tool usage:
>
> | Assisted-by: LLM [triage, tests], coccinelle
>
> I don't suggest this is included in this version, but it *is* fairly
> straightforward to add to AGENTS.md and tells reviewers just how exactly the
> LLM was used.

I think perhaps this should be a follow-up (this stuff is so fraught that I am
super-glad to see Christian's suggestion fly through and want to take gains
iteratively on this :), but completely agree.

David suggested something like:

Assisted-by: LLM # < what it actually did in the patch >

So I think we're kinda aligned on that?

Probably worth keeping the 'what it did' bit fuzzy.

>
> -K

--
Cheers, Lorenzo

Re: [PATCH] coding-assistants: simplify attribution

[David Hildenbrand (Arm)]

On 8/4/26 10:10, Lorenzo Stoakes (ARM) wrote:
> On Wed, Jul 22, 2026 at 04:48:03PM -0400, Konstantin Ryabitsev wrote:
>> On Wed, Jul 22, 2026 at 02:34:40PM -0600, Jonathan Corbet wrote:
>>> -  Assisted-by: Claude:claude-3-opus coccinelle sparse
>>> +  Assisted-by: LLM coccinelle sparse
>>
>> I've been playing with this a bit, and it's pretty easy to teach the agent to
>> add a few tag-like notes as to how exactly it was used, e.g.:
>>
>>     https://git.kernel.org/pub/scm/utils/b4/b4.git/commit/?id=018b87444d32eb2ffee7cb25f0785e54aa35f7bf
>>
>> This has:
>>
>> | Assisted-by: LLM [design, codegen, tests]
>>
>> It tells us a bit more about *how* the LLM was used, without introducing the
>> advertisement or pinning exact models.
>>
>> This format should allow for multiple tool usage:
>>
>> | Assisted-by: LLM [triage, tests], coccinelle
>>
>> I don't suggest this is included in this version, but it *is* fairly
>> straightforward to add to AGENTS.md and tells reviewers just how exactly the
>> LLM was used.
> 
> I think perhaps this should be a follow-up (this stuff is so fraught that I am
> super-glad to see Christian's suggestion fly through and want to take gains
> iteratively on this :), but completely agree.
> 
> David suggested something like:
> 
> Assisted-by: LLM # < what it actually did in the patch >
> 
> So I think we're kinda aligned on that?

Once the patch from Christian is upstream I can follow up with some
changes. I was holding that back for now until there is agreement on
the LLM part.

-- 
Cheers,

David

Re: [PATCH] coding-assistants: simplify attribution

[Jori Koolstra]

> Op 04-08-2026 10:10 CEST schreef Lorenzo Stoakes (ARM) <ljs@kernel.org>:
> 
>  
> On Wed, Jul 22, 2026 at 04:48:03PM -0400, Konstantin Ryabitsev wrote:
> > On Wed, Jul 22, 2026 at 02:34:40PM -0600, Jonathan Corbet wrote:
> > > -  Assisted-by: Claude:claude-3-opus coccinelle sparse
> > > +  Assisted-by: LLM coccinelle sparse
> >
> > I've been playing with this a bit, and it's pretty easy to teach the agent to
> > add a few tag-like notes as to how exactly it was used, e.g.:
> >
> >     https://git.kernel.org/pub/scm/utils/b4/b4.git/commit/?id=018b87444d32eb2ffee7cb25f0785e54aa35f7bf
> >
> > This has:
> >
> > | Assisted-by: LLM [design, codegen, tests]
> >
> > It tells us a bit more about *how* the LLM was used, without introducing the
> > advertisement or pinning exact models.
> >
> > This format should allow for multiple tool usage:
> >
> > | Assisted-by: LLM [triage, tests], coccinelle
> >
> > I don't suggest this is included in this version, but it *is* fairly
> > straightforward to add to AGENTS.md and tells reviewers just how exactly the
> > LLM was used.
> 
> I think perhaps this should be a follow-up (this stuff is so fraught that I am
> super-glad to see Christian's suggestion fly through and want to take gains
> iteratively on this :), but completely agree.
> 
> David suggested something like:
> 
> Assisted-by: LLM # < what it actually did in the patch >
> 

We should have some format to explain what the assistant was used for,
but not sure if a single line after the tag is going to cut it. I agree
it is the most simple, but I have personally used an LLM for a patch where
the explanation of what exactly it was put to do would have been very
hard to fit in a format like this without significant loss of information.

So, maybe below the ---? I guess that most maintainers don't want this kind
of explanation to be part of the commit message (or maybe they do?)

Best,
Jori.

Re: [PATCH] coding-assistants: simplify attribution

[Lorenzo Stoakes (ARM)]

On Tue, Aug 04, 2026 at 01:09:23PM +0200, Jori Koolstra wrote:
>
> > Op 04-08-2026 10:10 CEST schreef Lorenzo Stoakes (ARM) <ljs@kernel.org>:
> >
> >
> > On Wed, Jul 22, 2026 at 04:48:03PM -0400, Konstantin Ryabitsev wrote:
> > > On Wed, Jul 22, 2026 at 02:34:40PM -0600, Jonathan Corbet wrote:
> > > > -  Assisted-by: Claude:claude-3-opus coccinelle sparse
> > > > +  Assisted-by: LLM coccinelle sparse
> > >
> > > I've been playing with this a bit, and it's pretty easy to teach the agent to
> > > add a few tag-like notes as to how exactly it was used, e.g.:
> > >
> > >     https://git.kernel.org/pub/scm/utils/b4/b4.git/commit/?id=018b87444d32eb2ffee7cb25f0785e54aa35f7bf
> > >
> > > This has:
> > >
> > > | Assisted-by: LLM [design, codegen, tests]
> > >
> > > It tells us a bit more about *how* the LLM was used, without introducing the
> > > advertisement or pinning exact models.
> > >
> > > This format should allow for multiple tool usage:
> > >
> > > | Assisted-by: LLM [triage, tests], coccinelle
> > >
> > > I don't suggest this is included in this version, but it *is* fairly
> > > straightforward to add to AGENTS.md and tells reviewers just how exactly the
> > > LLM was used.
> >
> > I think perhaps this should be a follow-up (this stuff is so fraught that I am
> > super-glad to see Christian's suggestion fly through and want to take gains
> > iteratively on this :), but completely agree.
> >
> > David suggested something like:
> >
> > Assisted-by: LLM # < what it actually did in the patch >
> >
>
> We should have some format to explain what the assistant was used for,
> but not sure if a single line after the tag is going to cut it. I agree
> it is the most simple, but I have personally used an LLM for a patch where
> the explanation of what exactly it was put to do would have been very
> hard to fit in a format like this without significant loss of information.
>
> So, maybe below the ---? I guess that most maintainers don't want this kind
> of explanation to be part of the commit message (or maybe they do?)

I think best to keep discussion of this to the follow up patch to avoid the
usual combinatorial thread explosion on this topic :)

>
> Best,
> Jori.

--
Cheers, Lorenzo

Re: [tech-board] [PATCH] coding-assistants: simplify attribution

[David Hildenbrand (Arm)]

On 7/22/26 22:34, Jonathan Corbet via groups.linuxfoundation.org wrote:
> As suggested by Christian, I have applied his patch to a branch and
> simplified the changelog, see below.
> 
> I'm nervous about applying it still because I don't think it's my role
> to set this sort of policy; I really want to feel that there is a
> consensus behind it.  So there's some additional CCs here because I
> would like to hear from some folks:
> 
> - This policy document originally came out of the TAB; is the TAB (on
>   CC) OK with this change?
> 
> - At least one maintainer (on CC) have said that they currently strip
>   Assisted-by tags out of patches before applying them, a practice that
>   would seem to undermine the whole policy.  Would this change be
>   sufficient to end that practice?
> 
> Thanks,
> 
> jon
> 
> From c31aa376a355e3b84b9e87ae1dbe19b394d4c82c Mon Sep 17 00:00:00 2001
> From: Christian Brauner <brauner@kernel.org>
> Date: Wed, 1 Jul 2026 17:54:48 +0200
> Subject: [PATCH] coding-assistants: simplify attribution
> 
> The requirement to identify specific models used in the Assisted-by tag
> provides free advertising to proprietary software companies while adding
> little or no useful information.  Change the requirement to simply:
> 
>   Assisted-by: LLM
> 
> to capture the fact that an LLM was used without tracking which one.
> 
> Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
> Reviewed-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
> Reviewed-by: Jeff Layton <jlayton@kernel.org>
> Reviewed-by: Jori Koolstra <jkoolstra@xs4all.nl>
> Message-ID: <20260701-work-coding-assistants-v1-1-a20a94d1d606@kernel.org>
> Signed-off-by: Jonathan Corbet <corbet@lwn.net>
> ---
>  Documentation/process/coding-assistants.rst | 8 ++------
>  1 file changed, 2 insertions(+), 6 deletions(-)
> 
> diff --git a/Documentation/process/coding-assistants.rst b/Documentation/process/coding-assistants.rst
> index 899f4459c52d..fe34f3e7e828 100644
> --- a/Documentation/process/coding-assistants.rst
> +++ b/Documentation/process/coding-assistants.rst
> @@ -43,12 +43,8 @@ When AI tools contribute to kernel development, proper attribution
>  helps track the evolving role of AI in the development process.
>  Contributions should include an Assisted-by tag in the following format::
>  
> -  Assisted-by: AGENT_NAME:MODEL_VERSION [TOOL1] [TOOL2]
> +  Assisted-by: LLM [TOOL1] [TOOL2]
>  
> -Where:
> -
> -* ``AGENT_NAME`` is the name of the AI tool or framework
> -* ``MODEL_VERSION`` is the specific model version used
>  * ``[TOOL1] [TOOL2]`` are optional specialized analysis tools used
>    (e.g., coccinelle, sparse, smatch, clang-tidy)
>  
> @@ -56,4 +52,4 @@ Basic development tools (git, gcc, make, editors) should not be listed.
>  
>  Example::
>  
> -  Assisted-by: Claude:claude-3-opus coccinelle sparse
> +  Assisted-by: LLM coccinelle sparse

LGTM

Reviewed-by: David Hildenbrand (arm) <david@kernel.org>

-- 
Cheers,

David

Re: [PATCH] coding-assistants: simplify attribution

[Jonathan Corbet]

Jonathan Corbet <corbet@lwn.net> writes:

> As suggested by Christian, I have applied his patch to a branch and
> simplified the changelog, see below.
>
> I'm nervous about applying it still because I don't think it's my role
> to set this sort of policy; I really want to feel that there is a
> consensus behind it.  So there's some additional CCs here because I
> would like to hear from some folks:
>
> - This policy document originally came out of the TAB; is the TAB (on
>   CC) OK with this change?
>
> - At least one maintainer (on CC) have said that they currently strip
>   Assisted-by tags out of patches before applying them, a practice that
>   would seem to undermine the whole policy.  Would this change be
>   sufficient to end that practice?
>

So I was not overwhelmed by responses :)  I'm going to take silence for
assent in this case and send the patch through.

Thanks,

jon

Re: [tech-board] [PATCH] coding-assistants: simplify attribution

[Miguel Ojeda]

On Mon, Aug 3, 2026 at 9:42 PM Jonathan Corbet via
groups.linuxfoundation.org <corbet=lwn.net@groups.linuxfoundation.org>
wrote:
>
> So I was not overwhelmed by responses :)  I'm going to take silence for
> assent in this case and send the patch through.

On the TAB side, we didn't meet since it was sent, but yeah, it
doesn't look like there are strong opinions against this.

Acked-by: Miguel Ojeda <ojeda@kernel.org>

Personally, I think it is fine either way (e.g. someone could argue
the advertising could help open models and harnesses etc.).

The patch removes the "Where:", not sure if that is intentional.

Thanks!

Cheers,
Miguel

Re: [PATCH] coding-assistants: simplify attribution

[Jeff Layton]

On Wed, 2026-07-22 at 14:34 -0600, Jonathan Corbet wrote:
> As suggested by Christian, I have applied his patch to a branch and
> simplified the changelog, see below.
> 
> I'm nervous about applying it still because I don't think it's my role
> to set this sort of policy; I really want to feel that there is a
> consensus behind it.  So there's some additional CCs here because I
> would like to hear from some folks:
> 
> - This policy document originally came out of the TAB; is the TAB (on
>   CC) OK with this change?
> 
> - At least one maintainer (on CC) have said that they currently strip
>   Assisted-by tags out of patches before applying them, a practice that
>   would seem to undermine the whole policy.  Would this change be
>   sufficient to end that practice?
> 
> Thanks,
> 
> jon
> 
> From c31aa376a355e3b84b9e87ae1dbe19b394d4c82c Mon Sep 17 00:00:00 2001
> From: Christian Brauner <brauner@kernel.org>
> Date: Wed, 1 Jul 2026 17:54:48 +0200
> Subject: [PATCH] coding-assistants: simplify attribution
> 
> The requirement to identify specific models used in the Assisted-by tag
> provides free advertising to proprietary software companies while adding
> little or no useful information.  Change the requirement to simply:
> 
>   Assisted-by: LLM
> 
> to capture the fact that an LLM was used without tracking which one.
> 
> Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
> Reviewed-by: Lorenzo Stoakes (ARM) <ljs@kernel.org>
> Reviewed-by: Jeff Layton <jlayton@kernel.org>
> Reviewed-by: Jori Koolstra <jkoolstra@xs4all.nl>
> Message-ID: <20260701-work-coding-assistants-v1-1-a20a94d1d606@kernel.org>
> Signed-off-by: Jonathan Corbet <corbet@lwn.net>
> ---
>  Documentation/process/coding-assistants.rst | 8 ++------
>  1 file changed, 2 insertions(+), 6 deletions(-)
> 
> diff --git a/Documentation/process/coding-assistants.rst b/Documentation/process/coding-assistants.rst
> index 899f4459c52d..fe34f3e7e828 100644
> --- a/Documentation/process/coding-assistants.rst
> +++ b/Documentation/process/coding-assistants.rst
> @@ -43,12 +43,8 @@ When AI tools contribute to kernel development, proper attribution
>  helps track the evolving role of AI in the development process.
>  Contributions should include an Assisted-by tag in the following format::
>  
> -  Assisted-by: AGENT_NAME:MODEL_VERSION [TOOL1] [TOOL2]
> +  Assisted-by: LLM [TOOL1] [TOOL2]
>  
> -Where:
> -
> -* ``AGENT_NAME`` is the name of the AI tool or framework
> -* ``MODEL_VERSION`` is the specific model version used
>  * ``[TOOL1] [TOOL2]`` are optional specialized analysis tools used
>    (e.g., coccinelle, sparse, smatch, clang-tidy)
>  
> @@ -56,4 +52,4 @@ Basic development tools (git, gcc, make, editors) should not be listed.
>  
>  Example::
>  
> -  Assisted-by: Claude:claude-3-opus coccinelle sparse
> +  Assisted-by: LLM coccinelle sparse

One thing that needs to be changed in conjunction with this is
checkpatch.pl. When I try to use the new format, I get:

● a9b9b7e55379: btrfs: pre-allocate delayed dir index for non-overwrite rename
  ● checkpatch.pl: :2036: WARNING: Assisted-by expects 'AGENT_NAME:MODEL_VERSION [TOOL1] [TOOL2]' format

This patch fixes it for me, but I'm not sure it's restrictive enough.
Jon, feel free to fold this into the Documentation patch, or I can send
it separately if this seems right.

-----------------------8<-------------------------

[PATCH] checkpatch.pl: adapt to new Assisted-by: format

Signed-off-by: Jeff Layton <jlayton@kernel.org>
---
 scripts/checkpatch.pl | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/scripts/checkpatch.pl b/scripts/checkpatch.pl
index 2b7a42bbdd94..7961d9584bc5 100755
--- a/scripts/checkpatch.pl
+++ b/scripts/checkpatch.pl
@@ -3124,11 +3124,12 @@ sub process {
 				}
 			}
 
-			# Assisted-by uses AGENT_NAME:MODEL_VERSION format, not email
+			# Assisted-by uses a free-form value (e.g. "LLM"), not an
+			# email address, so skip the email format checks below.
 			if ($sign_off =~ /^Assisted-by:/i) {
-				if ($email !~ /^\S+:\S+/) {
+				if ($email =~ /^\s*$/) {
 					WARN("BAD_SIGN_OFF",
-					     "Assisted-by expects 'AGENT_NAME:MODEL_VERSION [TOOL1] [TOOL2]' format\n" . $herecurr);
+					     "Assisted-by requires a value\n" . $herecurr);
 				}
 				next;
 			}
-- 
2.55.0

Re: [PATCH] coding-assistants: simplify attribution

[Jonathan Corbet]

Jeff Layton <jlayton@kernel.org> writes:

> One thing that needs to be changed in conjunction with this is
> checkpatch.pl. When I try to use the new format, I get:
>
> ● a9b9b7e55379: btrfs: pre-allocate delayed dir index for non-overwrite rename
>   ● checkpatch.pl: :2036: WARNING: Assisted-by expects 'AGENT_NAME:MODEL_VERSION [TOOL1] [TOOL2]' format
>
> This patch fixes it for me, but I'm not sure it's restrictive enough.
> Jon, feel free to fold this into the Documentation patch, or I can send
> it separately if this seems right.

I'd say send it separately, probably with a copy to Joe Perches...

Thanks,

jon