

Most new code is written by LLMs nowadays. With the right human review, rules, and scrutiny, it works pretty well. It sounds like they are taking the right approach with a multi-stage pipeline, and it already benefits from having a patch ready-to-go from the CVE.
Of course, say anything like that on Lemmy, and everybody grabs their anti-AI pitchforks.




With a rather dumb counterargument, which is none at all.