@simplenomad@rigor-mortis.nmrc.org
Post #1822967
2026-04-21 15:03 UTC
@ryanjyoder Even just cutting and pasting code into Claude and asking it to find security flaws (obviously I use a much larger and detailed prompt) and give me a diff of fixes is nice. Now of course I don't blindly trust it, but after a lot of testing I can say that Claude is so far the best LLM for this. Besides us in security are the ones that are going to have to fix this crap whether we want to or not. So I'm seriously playing around with this stuff as there is some potential worth pursuing.
Replies (1)
-
@ryanjyoder@techhub.social 2026-04-21 15:49
@simplenomad It's definitely valuable. It's just a funny idea to me because I've already seen devs rerun workflows to get tests to pass. And the stochastic nature of LLMs will make them prime targets for lazy devs that just want to merge their code.