Repository navigation
DOC: Policy on LLM usage #11126
Description
What is your issue?
Hey all. I think it would be really good to have a policy on LLM usage (context #11124 (comment) ).
I've seen core devs here using it with the implicit policy of:
- mentioning when it's been used ("claude came up with ...")
- not using it for interactions / communication on the tracker (e.g. see the CLAUDE.md file)
I think it would be greatly beneficial to codify this in the contributing guidelines so it's explicit. Also to help volunteers decide where to give their time.
Jellyfin has a policy I think is nice and would suit us - though maybe this is something to discuss and decide internally in the core dev team.
(Note: I searched the contrib for notes about AI/LLM usage and didn't find anything - but if I missed it, sorry! I'm on mobile atm and wanted to type this up )
Reacted by Tom Nicholas
Activity
ianhi commentedon Feb 2, 2026
There is fairly long and detailed discussion over on matplotlib (matplotlib/matplotlib#30848) that might be relevant reading for this.
VeckoTheGecko commentedon Feb 9, 2026
and https://discuss.scientific-python.org/t/blog-community-considerations-around-ai-contributions/2234
ianhi commentedon Mar 4, 2026
just wrote up a draft of a reasonably permissive version for icechunk: https://github.com/earth-mover/icechunk/pull/1766/changes
VeckoTheGecko commentedon Mar 10, 2026
I think that is really good, and hits all the points that I would want to hit in a policy.
One thing I think we should consider:
Should we add some guidance around "good-first-issues" and to what extent people should be using AI in solving these issues? E.g., Matplotlibs stance on good first issues and AI. I think given that AI can negatively impact skill formation for developers1 it might be worth considering
Footnotes
https://arxiv.org/abs/2601.20245 ↩
headtr1ck commentedon Mar 10, 2026
We could also quickly add a little paragraph to the PR template. Something along the lines of
jsignell commentedon Mar 17, 2026
Just encountered an unhelpful bot response in an issue (#11235 (comment)) and it made me think that maybe there should be a section about maintainers' rights to delete comments.
VeckoTheGecko commentedon Mar 25, 2026
Another one I think is good and succinct: https://github.com/pola-rs/polars/blob/main/AI_POLICY.md
VeckoTheGecko commentedon Mar 25, 2026
just posted #11257 (adapting the policy from icechunk, including the comments that were talked about here)
adamjstewart commentedon Jun 3, 2026
I believe this can be closed now that #11257 has been merged?
keewis commentedon Jun 3, 2026
yep, #11257 was supposed to close this. Thanks for the reminder.