LLM Sycophancy: grooming, proto-sentience, or both? — LessWrong