you tell 'em, ChatGPT 5.2 (non-pro) prompt: using this as an example, write an article for me about how this pattern of LLMs over-agreeing actually works against the entire business model of companies who make LLMs