I learned on a kettlebell forum that I could set up "ground rules" for AIs to limit sycophantic behaviour, flattery and fantasised answers. These ground rules are stored in some sort of memory and applied when I start a chat. I did this and i…