2026-09-23 17:31 UTC
Sorry, but I think I'll pass. I don't need to know how good the models are at CTFs; we just had a bunch of news stories about that.
Replies (2)
-
@cR0w@infosec.exchange 2026-09-23 20:08
@mttaggart@infosec.exchange LMMs aside, how often do you get a chance to play a CtF directly with ransomware enablers?
-
@mttaggart@infosec.exchange 2026-09-23 21:57
Sorry, still annoyed about this, especially from someone I used to really look up to. The entire point of a CTF is to build skill—the very skill, by the way, necessary to verify any model output in a given field. Without knowing the stuff yourself, you can never know whether the model got it right. Making this about throwing tokens at a flag mistakes the flag for the value of a CTF, rather than the path that leads to the flag.