# GLINCKER

- **Event:** [Built with Opus 4.6: a Claude Code hackathon](https://cerebralvalley.ai/e/claude-code-hackathon)
- **When:** Feb 10 at 12:00 PM – Feb 17 at 10:00 AM (EST)
- **Where:** Location TBA
- **Team:** [GDS K S](https://cerebralvalley.ai/u/thegdsks)
- **GitHub:** https://github.com/thegdsks/verdict-engine-hackathon
- **Demo video:** https://youtu.be/LDc9VhOoJlE?si=2UCw6iBK0euO_l3l
- **Gallery:** https://cerebralvalley.ai/e/claude-code-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/claude-code-hackathon/hackathon/gallery/89

LLMs give one answer, one angle, zero pushback. Verdict spawns 3 to 5 AI agents who research, argue, and cross examine each other's claims. Every claim is tracked in an argument graph: standing, challenged, rebutted, or conceded. Opus 4.6 synthesizes a structured verdict with an emergent decision matrix where criteria come from the debate itself. Three novel algorithms: argument graph for claim lifecycle, protocol engine forcing direct engagement, and emergent decision matrix. Smart routing keeps costs under $0.72: Opus for synthesis, Sonnet for debate, Haiku for classification. 

A doctor tested it with a real medical question and got structured debate in 5 minutes where ChatGPT failed after 15 prompts. Ships as web app, npm SDK, CLI, and MCP server. Built solo in 6 days. Live at verdict.glinr.com, full cloud version launching soon.

---

Markdown version of https://cerebralvalley.ai/e/claude-code-hackathon/hackathon/gallery/89. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt