Post #1973051

2026-05-04 11:42 UTC

Curious question: When a PR comes in, how would you know its author abided by your policy? Personally, as a retired dev, I have not touched AI and can afford to continue to not touch it. But I do have a few low-traffic FLOSS projects out there (which see a PR once a year or even less often). So that's the background to my question. @blinry @teamtype

Replies (2)

@twilliability@genart.social 2026-05-04 15:21

@dj3ei @blinry This has been on my mind too. There are cases where a PR screams "AI generated", but it's a continuum from there to "author retyped an API call from a search's AI summary". It's hard because code in the end is just bytes, and their provenance is technically untraceable - like cash, and unlike some things in the physical world perhaps. (I'm not arguing for anything, just musing about the epistemology)

Open ##1973052
@blinry@chaos.social 2026-05-04 15:47

@dj3ei We have a "self-review" checklist in our PR template, which asks contributors to confirm that they have not used LLMs. I expect people to lie about this, though. Ultimately, there's no way we can be sure. But that doesn't mean we can't do our best to set our policy like we want.

Open ##1973053