On Sat, Jul 11, 2026 at 11:07 AM Jason Conroy <[email protected]> wrote: > > Greg Hogan <[email protected]> writes: > > > On Sat, Jul 11, 2026 at 3:50 AM Hugo Buddelmeijer via > > Development of > > GNU Guix and the GNU System distribution. <[email protected]> > > wrote: > > [...] > >> The GCD allows "exploratory analysis" using genAI [...] > > > > Au contraire, Hugo, not for you. Maybe. The author has stated > > his > > interpretation that the "will not use genAi" commitment takes > > precedence over the remainder of the document. The author has > > also > > stated the contrary, and even proposed wording that would make > > clear > > that the policy for contributions permissions are universal. > > I will assume that "will not use genAi" here is referring to Item > 1 under Project Commitments. Please note that there are important > qualifications after your selective quote. > > Prior discussion of the interaction between Project Committments > and Policy for Contributions has focused on the first two policy > items. If you feel that policy Item 4 (Exploratory analysis) is > also in confict, the considerate move would have been to offer > constructive feedback to the author directly - and it's not too > late to do so. > > As for Items 1 and 2 of Policy for Contributions: clearer wording > has been proposed [1], and the author has expressed support for it > [2], but you did not [3], [4]. As best I can tell, your objections > had nothing to do with clarity, but were because the clarified > text didn't match your preferred interpretation.
My objection was that your wording did not match the author's stated interpretation. > Expressing disagreement over positions and interpretations is > fine, but I object to the insinuation that any wording ambiguities > in the doc are evidence of blatant carelessness and lack of > motivation to reach consensus. This is not disagreement, rather the sections remain clearly in conflict despite our proposed changes. And if the author's intent is ambiguous then how can we propose and discuss changes to the document? I attempted several times to ground the discussion in relevant use cases which only lead to the author's contradictory statements. My stated preference was to segregate AI within or without the project (the latter an explicit ban, no policy exceptions). It's terribly unlikely that LLMs will be a little better than humans at this work. Either, as some have predicted, this will collapse into another AI winter or the remarkable progress will continue and make trivial building a non-Guix branded Guix with unlimited up-to-date, never broken, minimal closure packages. Barring a Chinese invasion or blockade of Taiwan, I expect that we are less than two years from this being feasible for a team and less than five years for an individual. In the interim all of our Guix boxes running out-of-date software will be exposed to newly discovered vulnerabilities revealed by soon commonplace Mythos-class models.
