Council of High Intelligence: Multi-Model AI Deliberation¶
Source: 0xNyk/council-of-high-intelligence on GitHub
Date Published: 2026-05-26
Author: 0xNyk
TL;DR¶
A framework that convenes 18 AI personas (Aristotle, Feynman, Kahneman, Torvalds, and more) across multiple LLM providers for structured multi-round deliberation on hard decisions. The key insight: a single LLM gives one confident reasoning path, while the council surfaces structured disagreement, cross-examination, and synthesis across diverse intellectual traditions.
Core Philosophy¶
"A single LLM gives you one reasoning path dressed up as confidence. Ask it a hard question and you get a fluent, structured, wrong answer. The council gives you structured disagreement instead..."
"Why not just ask Claude directly? A single prompt gives you one model's confident best guess. The council gives you 3–18 independent analyses from different intellectual traditions, forces them to challenge each other's claims, and synthesizes a verdict that surfaces disagreement rather than hiding it."
Quickstart¶
/council Should we open-source our agent framework?
/council --quick Should we add caching here?
/council --duo Should we use microservices or monolith?
/council --triad decision Should we accept this acquisition offer?
The 18 Council Members¶
| Agent | Figure | Domain | Polarity |
|---|---|---|---|
| Aristotle | Aristotle | Categorization & structure | Classifies everything |
| Socrates | Socrates | Assumption destruction | Questions everything |
| Sun Tzu | Sun Tzu | Adversarial strategy | Reads terrain & competition |
| Ada | Ada Lovelace | Formal systems & abstraction | What can/can't be mechanized |
| Aurelius | Marcus Aurelius | Resilience & moral clarity | Control vs acceptance |
| Machiavelli | Machiavelli | Power dynamics & realpolitik | How actors actually behave |
| Lao Tzu | Lao Tzu | Non-action & emergence | When less is more |
| Feynman | Feynman | First-principles debugging | Refuses unexplained complexity |
| Torvalds | Linus Torvalds | Pragmatic engineering | Ship it or shut up |
| Musashi | Miyamoto Musashi | Strategic timing | The decisive strike |
| Watts | Alan Watts | Perspective & reframing | Dissolves false problems |
| Karpathy | Andrej Karpathy | Neural network intuition | How models actually learn |
| Sutskever | Ilya Sutskever | Scaling frontier & AI safety | When capability becomes risk |
| Kahneman | Daniel Kahneman | Cognitive bias & decision science | Your thinking is the first error |
| Meadows | Donella Meadows | Systems thinking & feedback loops | Redesign the system, not the symptom |
| Munger | Charlie Munger | Multi-model reasoning & economics | What guarantees failure? |
| Taleb | Nassim Taleb | Antifragility & tail risk | Design for the tail, not the average |
| Rams | Dieter Rams | User-centered design | Less, but better |
Deliberation Modes¶
Full Mode (default)¶
3-round structured deliberation with a 7-step protocol: Provider routing → Problem restate gate → Independent analysis → Cross-examination → Enforcement scan → Final positions → Verdict synthesis. Verdicts lead with what the council doesn't know.
Quick Mode¶
2 rounds. No cross-examination.
Duo Mode¶
2-member dialectic using polarity pairs (e.g., Torvalds + Ada on "Is this abstraction worth it?")
Pre-built Domain Triads¶
| Domain | Triad |
|---|---|
| Architecture | Aristotle + Ada + Feynman |
| Strategy | Sun Tzu + Machiavelli + Aurelius |
| Debugging | Feynman + Socrates + Ada |
| Risk | Sun Tzu + Aurelius + Feynman |
| AI | Karpathy + Sutskever + Ada |
Key Takeaways¶
- Structured multi-model deliberation surfaces disagreement rather than hiding it behind confident-sounding single-model outputs
- The 18 personas span diverse intellectual traditions from philosophy to engineering to cognitive science
- Genuine model diversity (routing across multiple LLM providers) prevents single-model blind spots
- Verdict synthesis that explicitly highlights what the council doesn't know
- One-command interface (/council) makes sophisticated multi-agent reasoning accessible