Re: GR: Ban LLM contributions from Debian

Pierre-Elliott Bécue <[email protected]> Thu, 23 Jul 2026 17:53:26 +0200

Newsgroups	gmane.linux.debian.devel.vote
Message-ID	<[email protected]>

Ansgar 🙀 <[email protected]> wrote on 23/07/2026 at 16:41:08+0200:

> Hi,
>
> On Thu, 2026-07-23 at 16:28 +0200, Pierre-Elliott Bécue wrote:
>> > Yes, and Debian extracts work from many "repo from which work can be
>> > extracted into a distro" that are not under Debian's control and for
>> > which Debian cannot enforce/implement policies.
>> > 
>> > What is the difference of a LLM pushing into, for example, the python
>> > upstream repository vs. a LLM pushing into a packaging repository for
>> > python (which also might already include the work the former LLM
>> > pushed)?
>> > 
>> > Ansgar
>> 
>> We are free to have a higher standard for ourselves that others do.
>
> Given we don't have mandatory code reviews, we certainly don't do that.

Which doesn't mean we shouldn't aim to that, especially with generative
AI production.

>> The fact that upstream do have lower standards and do accept their
>> agents to effectively push potentially problematic code or content
>> into their repos is their right.
>
> The claim that upstreams have lower standards doesn't seem warranted.

To me, allowing a generative AI agent/model to push code on its own is a
lower standard that what I would expect.

> I've seen enough bad code in Debian.

Are you stating that "because some people are not doing a great job, we
should lower the bar so that even them fulfill our expectations"?

-- 
PEB

signature.asc (application/pgp-signature, 853 B)