Oh, that’s an interesting thought, I was really hoping LLMs would break the cycle there but of course there’s no reason to assume they’d be immune to adversarial content optimization.