We've just rolled out a new policy for using generative AI tools when contributing to BeeWare.
In short, we don't care if you use AI tooling or not. However, we *do* expect that a human takes responsibility for everything that is contributed - including adhering to project guidelines - and for that human to declare if they've used AI tooling.
The full policy can be found here:
https://beeware.org/contributing/guide/policies/ai-policy/
@beeware I did, of course. I am not a BeeWare maintainer, so I don't know how strict the code reviewers are, and whether they are then assumed themselves to understand the contributed code long-term.
My concern is that allowing code from a source that is known to be harder to maintain is quite different than accepting it from humans who have a much higher likelihood of becoming long-term maintainers themselves.
@paulehoffman The policy says, explicitly, that:
* A human is always responsible
* The human contributor must understand the code they contribute
* The human reviewer must understand the code they're reviewing.
There is functionally no difference between accepting a PR from a drive-by human contributor, and accepting an AI-generated PR.
Yes, there is a different long term load balance on the maintainers, but that's an entirely different issue, and one unlikely to be fixed with a policy.
BeeWare Project
Paul Hoffman