What's this about?
Debian's general resolution on large language models has been decided, and choice 5 won. The text: Debian neither endorses nor prohibits generative AI tools in development, maintenance or documentation. It recognises the tools can substantially improve contributor productivity when used responsibly, letting volunteers spend more of their limited time on work needing expertise, judgment, review and collaboration. And then the operative clause: all contributions, regardless of how they were produced, must meet the same standards of quality, correctness, maintainability and legal compliance — and using a generative AI tool does not diminish the contributor's responsibility for what they submit.
What is a Debian general resolution?
Debian is governed by its developers rather than a company or foundation board. When a question is contentious enough that no maintainer can settle it, any developer can put it to a project-wide vote using Condorcet ranked ballots — you rank all options including "none of the above", and the winner is the one that beats every other option head-to-head. It is slow, produces genuinely legitimate outcomes, and is why a Debian position carries weight that a corporate policy does not.
What's the community saying?
- The most common first reaction is a Windows 11 comparison, which turns into a long argument about whether Microsoft's recent problems are attributable to generative AI at all.
- The more considered counter-position: it is more plausible that leadership lost the plot, and there is a long history of that predating LLMs.
- A recurring structural theory — repeated salary-based cuts remove the experienced people who understood core code, leaving juniors maintaining it — which several people find a better explanation than tooling.
- Worth noting how little of the thread engages with the actual resolution text, which is unusually well-drafted: it moves the question from "was AI used" to "is the contributor accountable", which is the only enforceable version.