Post #2475819
2026-04-23 13:04 UTC
@ianthetechie@fosstodon.org That was pretty much the only point I disagreed with. But it's a give and take with these kinds of organization wide policies.
What I think is most useful to know is which part of a PR was written with the help of AI. Because that parts needs the most skeptical eyes.
Also I think prompts may help us retrace our steps and regenerate or modify parts of the same code in the future.
Lastly, some models produce better or worse results than others.
Replies (1)
-
@ianthetechie@fosstodon.org 2026-04-24 02:21
@bart@floss.social Ok; good to know :) (I don't really have any "problem" with such a policy; I just don't understand the point haha). > Also I think prompts may help us retrace our steps and regenerate or modify parts of the same code in the future. I'm not so sure about this one 😂 Given that even the same model with a slightly different harness version can produce something quite different. To be clear, the nondeterminism is a feature, but idk if this actually helps "debug" or "retrace" much.