Open Council
A multi-model debate orchestrator: several AI models debate one question, peer-review each other, and reach a measured consensus for an answer more reliable than any single model. Low consensus means no forced conclusion. Orchestration and full transcripts stay on your machine; models can be official APIs or a local Ollama / vLLM endpoint. MIT.
5 postsAll posts →
- Make the AIs Hold a Meeting: Open-Sourcing Open Council
- Cutting OAuth, Keeping Two Protocols: The Trade-offs of an Access Convergence
- The Debate Is Too Long to Feed In: A Compression Before Synthesis
- Every Chinese Keyword Was Dead Code: Routing, Roles, and a Regex Trap
- How a Consensus Score of 0.82 Is Computed: Quantifying Multi-Model Peer Review