Tolani Omitokun
The Emergent Diagnostic Institution is a six-agent AI system built with Claude Opus 4.6 and Claude Code.
Core trick
Three specialist agents debate a case across rounds while an observer agent flags cognitive biases, and a rules file is rewritten after each case.
Description
The Emergent Diagnostic Institution is a six-agent AI system built with Claude Opus 4.6 and Claude Code. Three specialists independently analyze complete medical histories, then engage in multi-round debate. A Metacognitive Observer detects cognitive biases like anchoring and premature closure. A Patient Translator writes plain-language explanations for families. A Constitution Amender rewrites the system's rules after every case. It processes 77,000+ words of medical records (not summaries), agents challenge each other through structured disagreement, and the system evolves after every case, generating 43 amendments across 6 evaluations. The Observer caught 24 cognitive biases. In the showcase case, EDI analyzes a boy misdiagnosed with autism for 12 years across 8 doctors and 3 states, identifying CLN2 Batten Disease with 91% confidence. 100% accuracy across 6 domains. Built in 3 days, solo. Live: https://emergentdiagnostic-ai.up.railway.app (Code: DEMO2026)
tolaniomitokun/emergent-diagnostic-institution★ 00 forksJavaScriptMITCreated Feb 16, 2026 · last push Feb 24, 2026
README excerpt
# The Emergent Diagnostic Institution A self-governing AI diagnostic institution where six specialized agents debate complex medical cases, detect their own cognitive biases in real-time, and evolve their clinical reasoning constitution after every case. **[Live Demo](https://emergentdiagnostic-ai.up.railway.app)** | **[Demo Video](https://youtu.be/your-video-id)** Built with Claude Opus 4.6 and Claude Code for the **"Built with Opus 4.6" Hackathon** (Anthropic/Cerebral Valley, Feb 2026). --- ## What It Does Give the institution a patient case — paste clinical notes, upload medical records, or describe symptoms — and it runs a multi-agent diagnostic pipeline: 1. **Specialist agents** independently analyze the case through their domain expertise 2. **The Metacognitive Observer** reviews all analyses for cognitive biases (anchoring, premature closure, confirmation bias, diagnostic momentum) 3. **Specialists debate** in a second round, forced to address the biases the Observer detected 4. **The Chief Diagnostician** synthesizes everything into a final institutional diagnosis with calibrated confidence 5. **The Patient Translator** writes a plain-language explanation for the patient's family 6. **The Constitution Amender** proposes new institutional rules based on lessons learned The institution gets smarter after every case. ## Showcase Case: Eli Reeves A 14-year-old boy misdiagnosed with autism for 6 years while actually having CLN2 Disease (a treatable neurodegenerat