On Sun, Jul 5, 2026 at 9:36 PM Ludovic Courtès <[email protected]> wrote: > > Hi Greg, > > Greg Hogan <[email protected]> skribis: > > > Since we recognize that this is an imperfect policy the only path > > available to amend the document is to add a sunset or expiration > > clause. A period of one year would be reasonable and allow time for > > the publication of GNU guidelines as well as for project contributors > > to explore the use of generative AI in the limited cases outlined by > > the GCD. > > I am somewhat reluctant to automated expiration because that could leave > us in limbo if the measures chosen in the GCD expire and nobody picks up > the work of consulting the community to determine what to do next.
GCD 008 will update the project documentation, pull request templates, etc. None of that expires. If after a year the community is satisfied such that the sunset date passes unnoticed then we continue under the GCD 008 policies. > It seems safer to me to come up with a GCD defining a new policy > whenever we have input that calls for a change. It forces us to have a > discussion on how to address these changes. The GCD process requires 100% consensus for every change. > > As to the current document, I still find unclear the project > > commitment to "will not use genAI" with the separate policy for > > contributions which documents the use thereof. > > As I tried to explain, this is pragmatic: if a newcomer comes up and > proposes a “simple” patch claimed to be LLM-assisted, I would rather not > turn them down. > > But the project itself, people who are explicitly engaged as team > members etc., can make a stronger commitment. That’s the spirit. How does this comport with 'In my mind “contributors” in the policy is any contributor, including “project members”.' [0]? [0] https://lists.gnu.org/archive/html/guix-devel/2026-06/msg00218.html > > And with the recent change under "contribution acceptance" in > > bd04d0962a I would be happy to see included something similar to > > "mechanical integration of upstream bug fixes, configuration > > parameters, and minimal source adaptations required to achieve a > > successful Guix build". > > I wouldn’t want the list of “Examples of non-creative changes” under > “Contribution acceptance” to be too long because they are just that, > examples. Do you see any ambiguity here? I find more examples to be less ambiguous than fewer examples. The document should be as clear as possible, otherwise the consensus is of no value if everyone has their own understanding. There was another similar, small change in eac798a510 that deleted "translations" when translations are still included in "any other artifact". How was this clarifying? The change in bd04d0962a swapped the reference to "legally significant" for a ban on all copyrightable contributions. This is a significant change so late in the discussion period! I do not see consensus for a total ban on generative AI for the community, and a near total ban for external contributors. Note that the total ban includes the use of generative AI for exploratory analysis. Most will not even know when they are using AI. Web searches now use generative AI, so searching for a build error or bug will now be a Guix violation.
