02:17:52 <br-m> <ack-j:matrix.org> Hi yes that was found by opus 4.8 or opus 5. I cant remember > <@rucknium> If I may ask, did an LLM assist with finding @ack-j:matrix.org 's bug?
02:51:37 <br-m> <ack-j:matrix.org> (Not a formal offer but gauging interest) Would devs be interested in MAGIC sponsoring the cost of having claude review Github PRs?
02:55:08 <br-m> <ack-j:matrix.org> If yes:
02:55:08 <br-m> <ack-j:matrix.org> 1. I believe it would be most effective to limit who could initiate claude to perform a review (restricted to core devs so we dont waste credits on typo prs for example)
02:55:08 <br-m> <ack-j:matrix.org> 2. The claude review should not count towards the 2 mandatory reviews needed to merge.
02:55:49 <br-m> <ravfx:xmr.mx> Each time a new PR get merged, it shoud audit the whole project (that's just what I think because it's probably what the adversary does... But they have virtually unlimited tokens).
02:56:42 <br-m> <ravfx:xmr.mx> A new PR can trigger a bug somewhere else.
02:58:25 <br-m> <ack-j:matrix.org> I disagree, that would cost a lot of money and produce worse results. We can do audits of the codebase outside of CI/CD. The llm is smart enough to look at the downstream effects of the code change
02:58:55 <br-m> <ravfx:xmr.mx> Yes, it would cost a lot of money.
03:03:02 <br-m> <jpk68:matrix.org> @ack-j:matrix.org: I can barely be considered a "dev" (if at all), but I don't think that's a good idea. It would cost way too much and mostly be unhelpful
06:58:44 <br-m> <loop.ster:matrix.org> If you wanted to go that route, CodeRabbit is already free for OSS projects