Questions about binutils AI policy?

[Mark Wielaard]

Hi Nick,

On Fri, Jan 23, 2026 at 10:50:03AM +0000, Nick Clifton wrote:
> >>Should the discussion about markup be moved to the Footnotes
> >>for a future policy discussion?
>
> Agreed, that does make more sense.
> 
> So I have taken the moved the offending paragraph to the footnotes
> and reworded it to say:
> 
>   When submitting a non-legally-significant LLM generated change,
>   it is encouraged to clearly indicate the use of the LLM. The
>   identification could take the form of a line starting with the
>   Generated-By: prefix which identifies the LLM used. For example:
> 
>     Generated-By: GNU-LLM version 1.0
> 
>   The reason for asking for this is to set a precedent.  So in
>   the future, if non-trivial LLM generated patches do become
>   acceptable, the process of labeling them will already be a
>   standard action.
> 
> Does that sound OK ?

Given that it is in a footnote and isn't part of the policy just an
idea for a discussion that should be fine. As long as people know it
isn't mandatory and might not be very helpful.

I don't agree suggesting this "generated-by" for trivial code snipets
is very useful. Since it is non-legally-significant you wouldn't do
that so formally for any other case. We also don't say
Single-Line-Copied-From-Paper: [DOI] Digital-Object-Identifier. But we
would describe the idea from the paper that inspired the code
generally in the commit message.

Also, but that is of course part of the discussion, I don't think such
a simple oneline tag can describe the input/output/model interaction
that helps others figure out how to regenerate and verify what an LLM
produces.

So I would start the footnote with something like: "If in the future
there are significant changes to how LLMs operate and using their
output is less contriversal we might accept them with a policy like
..."

Cheers,

Mark