Having One AI Roast Another: User Gets Model to 'Review' Gemini With Hilarious Results
zsakib_ · x · 2026-10-01
A user shared a playful experiment: asking a model to write a critical 'review' of Gemini and letting it draw its own conclusions, calling the outcome scandalous and hoping Claude wasn't too mean. Cross-vendor model roasts like this circulate as community fun rather than rigorous evaluation.
Related event: User Has Claude Opus 5.5 Review Gemini 4, Results Are Dramatic(2 posts)→
More from Fun
- Gemini 4 Argon missed #1 on Vending-Bench due to memory slip on test end date — infoxiao · 2026-10-01
- AI speeds up research so much you can answer an arXiv question the same day — KyleCranmer · 2026-10-01
- User reports being unable to cancel ChatGPT subscription — imjustnewatai · 2026-10-01
- Polish netizen jokes Poland was ahead on AI: it's always been 'SI' in Polish — tlakomy · 2026-10-01
- OpenAI researcher jokes model release finally excuses his late replies to friends and family — divy93t · 2026-10-01
- Skip the AI "muse" — this user wants a scheming, silver-tongued vizier persona — tedmitew · 2026-10-01