This prompt trick forces AI to stop flattering you and think harder
… But even as LLM providers have become aware of AI sycophancy and are training them to be more critical, it's still easy to get an AI to enthusiastically endorse a shaky theory that doesn't deserve it. …