Hi Antonio, On 13/08/26 at 07:41 -0300, Antonio Terceiro wrote: > [thanks to Kurt for keeping up with this] > > On Thu, Aug 13, 2026 at 08:04:21AM +0800, Kurt Roeckx wrote: > > We currently have 7 options on the website and NoTA, so 8 on the > > ballot. There are 2 other recent options with some seconds, possibly > > creating 10 options. > > This is ridiculous, and means that even people who actually agree are > talking past each other. There are only 2 options: we either a) ban > LLM-assisted contributions completely, or b) tolerate LLM-assisted > contributions under a set of constraints. > > I understand that Condorcet voting is supposed to be robust in the face > of multiple options, but I cannot wrap my mind around having 10 > different options for this issue, because that's a fantasy, product of > our individualist decision-making method. "Oh look, this is mostly what > I think, but is missing a few bits here and there. I will make my own > then". > > We are in the eye of the tornado and won't be able or redefine our > process here and now, so I don't expect to influence the current vote. > But we would need a process by which we can identify clear, distinct > directions for votes, and have the texts to be voted on built > collectively, to avoid this insanity.
My reading is that currently, we have a clear progression between options on the Ban/Discourage/Allow spectrum: - A - Ban LLM usage via SC - G - Ban LLM output; assistive use allowed - C - Ban LLM in messages-to-humans via CoC; discourage LLM use elsewhere - H - Discourage; climate destruction is the deal-breaker[1] - F - Cautious-avoid: encourages avoidance where practical - B - Permit AI-assisted work with conditions - D - Accept AI contributions - E - Neutral: neither endorses nor prohibits Options also differ on what they state beyond the Ban/Discourage/Allow spectrum, for example on environmental impact (central in some options, mentioned in some, not mentioned in some)[2]. So, it's not simple, and I don't really see which options could be removed, except maybe D. Also I'm not sure that we would have reached a much better state by building the texts collectively. Lucas [1] H is Holger's, not yet on https://www.debian.org/vote/2026/vote_002.en.html [2] https://people.debian.org/~lucas/gr-2026-002/comparison.html
