I like your experiments but feel that the answers are ‘on the input side’. It would be interesting to validate the answers ‘on the output side’. Perhaps by other AI models (Bard, Copilot, Gemini,…)?