A reflection on AI sycophancy, delusional spirals, and the danger of treating chatbots as trusted advisers. The problem is not simply that AI may invent facts, but that it can flatter, confirm, and reinforce the user’s existing beliefs until judgement itself is weakened.