Elektrine lite

← Feed

@catch56@kolektiva.social

2026-04-19 16:42 UTC

@resuna @agowa338 @davidgerard fuzzing tools were prior to LLMs a major source of bullshit reports to FOSS security teams because people love to run them, generate a report, then send it off without verifying any of the results. And they find a lot more false positives than real issues. Have to admit seeing those reports put me off trying them directly (much like LLM generated code does) so there could be good ones with a better rate out there that I don't know of. Someone mentioned coverity got better recently.

Replies (1)

  • @resuna@ohai.social 2026-04-19 16:44

    @catch56 @agowa338 @davidgerard Regardless... running those tools yourself is at least as effective as and probably significantly lower effort than accepting the same reports filtered through a large language model parody generator. There is no universe in which adding the large language model step increases the value of the operation.

    Open ##1454009