This is brilliant :-) thank you for this. I'd never come up with telling a LLM to "quietly think"... Now I'll be testing this with all my OS models.