>> My understanding of the AI policy is that changes outside of the testsuite > need to be authored entirely by a human (unless the patch is trivial).
I don't think the policy was that clear. > using AI to write the fix is out of bounds. I think that's clear. > In other words, it's ok (AIUI) for someone to use AI to detect flaws and > create testcases for those flaws ... and that's clear. > and for that person to then go off and write a fix without the help > of AI. I think that's not really addressed by the policy and I think is too strict. Let's suppose you've done the above and an AI model identified a problem. If you next ask it "how do you recommend fixing it?" and it gives a general description of what to do and you do it, I don't see any potential problem. For example, if it says "I'd write a predicate that tests XYZ and call it here" and you do it, the code was still human-written.
