2026-09-09 00:58 UTC
I think its ALSO super harmful to die on this particular hill bc I'm reading the recent Graphene OS PRs and it IS as they say.
I see vigiours HUMAN discussion on most PRs. I'm not seeing tons of tell-tale LLM usage on the code generation side.
I DO see evidence of LLM usage for security review... which imo... you're gonna get wrecked rn if you're not doing LLM reviews there is a legit arms race on rn.
@dalias@hachyderm.io
Replies (1)
-
@dalias@hachyderm.io 2026-09-09 01:02
@nullagent@partyon.xyz I don't think that's at all a serious position. Yes in their position you probably want someone running whatever heuristic vuln scanning utilities attackers are using, even if some of those are LLM-slop. You do NOT need to be actually using "LLM code review" from those things as part of your development process. You can take a list of potential issues, condensed down to 1 line descriptions with a pointer to the line/function in your code, and actually review them like you would anything out of a static analysis tool.