Yes separating the two is my goal. Same with testing standards. In hopes we can drive some consensus on AI guiding principles and then tackle those discussions separately, perhaps in parallel.
I think as proposed what I wrote is far from a blanket ban. It intentionally uses language that would hopefully make a contributor think about their LLM usage in this context but not feel like the decision was made for them. But again also happy to change that wording to something else to better reflect that. Jordan On Wed, Sep 30, 2026 at 09:31 Josh McKenzie <[email protected]> wrote: > I would make the case that’s a project level standard and maybe the LLM > guidelines doc then states something to the effect of “LLM generated docs > and comments must meet [linked or insert guidelines]. As with code, quality > of documentation and comments is the responsibility of the contributor”. > I’m not sure all community members here will feel comfortable with that > given the ongoing discussion but I’m happy to support it myself. > > Good distinction; this is kind of a 2 part problem. We both a) don't have > clear definitions we agree upon for what quality comments or docs look like > (that I can think of or find), so we b) can't say "here's the standard. > Live up to it with or without an LLM and you're accountable for the quality > of what you contribute." consistent with the rest of the guidelines. > > I don't want to conflate the two and gum up the works on a proposal like > this, but I'm not ok with us blanket banning usage of *any* tools when I > think the right thing is having a bar of quality and care we agree to hit > and leave it up to contributors to decide how best to hit it. Same thing > w/docs and comments as with tests and code, etc. > > On Wed, Sep 30, 2026, at 10:43 AM, Brandon Williams wrote: > > On Tue, Sep 29, 2026 at 2:30 PM Jordan West <[email protected]> wrote: > > * All LLM use must be attributed to both you and the harness, provider, > and model being used. The project makes no specific recommendations or > requirements regarding the toolchain used as long as the user has legal > access and provides attribution. > > I think recording this in every commit and thus making it publicly > available provides more negatives than positives. AI is both heavily > marketed and politicized, requiring this data be made available leaves > the project vulnerable to being used by either in the future, whatever > that brings. > > Kind Regards, > Brandon > >
