What if you use OpenAI as a judge and check both responses? Not defending Gemini, just curious if it was that wrong, or Fable - that aggressive.