That's not what i'm talking about though. The point is that for spam detection, LLMs are up against other probabilistic measures. No-one sane is detecting spam

That's not what i'm talking about though. The point is that for spam detection, LLMs are up against other probabilistic measures. No-one sane is detecting spam with if-then's. You simply do not have the luxury of rigid reliability.