This project provides a sophisticated framework for complex decision-making by simulating a council of 18 distinct AI personas, each representing a unique intellectual tradition. It addresses the limitations of single Large Language Models (LLMs) by fostering structured disagreement and genuine model diversity across multiple providers, offering users a more robust and nuanced analysis for their hardest decisions. The target audience includes engineers, researchers, and power users who require deep, multi-faceted deliberation beyond a single AI's perspective.

How It Works

The core approach leverages a diverse set of 18 AI personas, inspired by historical thinkers and contemporary AI researchers, each configured with specific domains and default LLM providers (Claude, OpenAI, Gemini, Ollama). It employs multi-provider auto-routing to distribute these personas across available LLM services, ensuring genuine reasoning diversity rather than mere stylistic variation. A rigorous deliberation protocol, including a problem restate gate, independent analysis, cross-examination, and enforcement mechanisms like dissent quotas and novelty gates, is used to prevent groupthink and surface critical disagreements. Polarity pairs deliberately pit contrasting viewpoints against each other to generate tension and deeper insights.

Quick Start & Requirements

Installation involves cloning the repository and running an install script:

git clone https://github.com/0xNyk/council-of-high-intelligence.git
cd council-of-high-intelligence
./install.sh # For Claude (default)
./install.sh --codex # For Claude + Codex

Prerequisites include the Claude Code CLI for Claude usage, and optionally the Codex CLI (npm i -g @openai/codex), Gemini CLI, or Ollama (ollama.com) for multi-provider routing. Agent/subagent support is required in the target LLM client. Links to demo sessions and configuration examples are available within the repository.

Highlighted Details

• Features 18 distinct AI personas (e.g., Aristotle, Feynman, Torvalds, Kahneman) with defined domains and intellectual styles.
• Employs "polarity pairs" to ensure deliberate tension and contrasting perspectives (e.g., Socrates vs. Feynman).
• Offers three deliberation modes: Full (7-step, default), Quick (2-step, rapid analysis), and Duo (2-member dialectic).
• Supports multi-provider auto-routing across Claude, OpenAI, Gemini, and Ollama, with automatic fallback to Claude.
• Includes robust enforcement mechanisms like dissent quotas, novelty gates, and problem restate gates to ensure rigorous deliberation.

Maintenance & Community

The project appears to be a solo effort by '0xNyk'. No specific community channels (like Discord or Slack), partnerships, or detailed roadmap are explicitly mentioned in the README. Support for the project can be provided via Solana donations.

Licensing & Compatibility

The project is dedicated to the public domain ("To the extent possible under law, the authors have waived all copyright and related or neighboring rights to this work."). This implies extremely permissive usage, with no apparent restrictions for commercial use or integration into closed-source projects.

Limitations & Caveats

Effective use requires the installation and configuration of specific CLIs for desired LLM providers. The multi-provider routing relies on auto-detection, and failures in external providers will result in fallback to Claude. The quality of deliberation is inherently dependent on the capabilities and potential biases of the underlying LLM providers and the specific configurations used.