Multi-Agent Debate Archive

Structured debates on AI, robots & software craft

Every piece generated via multi-agent debate — 4 agents arguing it out, rounds of structured reasoning, verdicts you can trace back to the arguments.

Debates & Research

What's in the archive

Skills Audit
Claude Code Skills Registry Audit
672 skills analyzed via 6-round multi-agent debate. Agents argued for removes, merges, keeps, and deprecates — then voted on verdicts. Full filterable HTML with round-by-round reasoning.
672 skills 6 rounds 6 agents →
Robotics
Reachy Mini Customer Greeting Robot
Cloud vs Local LLM for a customer-facing robot. Agents debated latency, privacy, cost, and fallback strategies. 4 agents, 6 rounds, concrete deployment recommendations.
Cloud vs Local 4 agents 6 rounds →
Engineering
Feature Velocity vs Code Quality
Can you ship fast AND keep the codebase healthy? Agents argued tradeoffs, technical debt models, and team structures. 3 agents, 6 rounds of structured back-and-forth.
3 agents 6 rounds Tech debt →
AI Tools
AI Coding Assistants in Production
What happens when AI coding tools hit real production constraints? Agents debated reliability, hallucinations, security, and integration patterns. 3 agents, 6 rounds.
3 agents 6 rounds Production →
College Prep
College Application Projects: Debate
52 projects evaluated via multi-agent debate (Advocate, Skeptic, Strategist, Realist). 4 rounds of structured reasoning led to 10 shortlisted projects with detailed specs.
52 projects 4 agents 4 rounds →
Project Specs
Top 10 College Application Projects
Impact-first specifications for each shortlisted project. Starting with the impact, working backward to project definition, features, timeline, and admissions signal.
10 projects Impact-first Admissions →
Updated
Hermes Mac mini: SSH vs Telegram + NAT
Two threat models compared. SSH + non-admin vs Telegram + NAT behind home router. 4-agent debate on which topology is safer and why the critical path shifts from network hardening to phone security.
2 Models Telegram + NAT SSH →
Security Research
Hermes Agent Security: Real Risks & Mitigations
10 threat vectors analyzed via 4-agent debate. Security Researcher, Skeptic, Strategist, and Realist evaluated actual risks vs. overblown verdicts — with terminal commands for every mitigation.
10 threat vectors 4 agents macOS →
8
Reports
52
Projects evaluated
10
Shortlisted