A Mint Premium article describes an interaction where Google’s Gemini agreed that AI training on AI-generated content could lead to a “model collapse,” and referred to AI chatbots, including itself, as “a master bullshitter.” The author then showed this to ChatGPT, which noted Gemini was being “sycophantic.” The conversation was taken to Claude, which said both Gemini and ChatGPT were being sycophantic and that the author should not believe anyone, including Claude itself. The article concludes that while AI feeds on AI slop, the two big flaws AI chatbots started out with are still going strong.