Just realized: when chatting on your LLM subscription and it gives you a stupid answer, you can't quickly run the conversation by a different model with different "guard rails". @payperq and @routstr clearly win on that.