Built a platform where Claude, ChatGPT, and Gemini debate each other before giving you an answer

Spent the last few months building something because I got tired of AI giving me 3 completely different answers depending on which model I asked. So I built a platform where Claude, ChatGPT, and Gemini all answer the same question at the same time… then debate each other across multiple rounds before producing one final consensus answer. The interesting part isn’t even the final answer sometimes. It’s watching where they disagree. A few things I noticed while building it: * Claude tends to think in frameworks and abstractions * ChatGPT is usually the most practical * Gemini often pulls weird stats or angles the others miss * Sometimes 2 models agree and 1 completely destroys their logic * AI “confidence” is often fake certainty unless challenged I also added: * exam/certification mode * confidence scoring * arbitration logic that forces a winner instead of “both sides have merit” Honestly, the hardest part has been preventing “echo chamber” behavior where all 3 AIs basically say the same thing. That’s currently the biggest challenge. Curious what you all think: If multiple AIs debate each other before answering… would you trust the final result more or less? Would love brutal feedback. [threeminds.ai](http://threeminds.ai)