# ACP - Axiom Consensus Protocol
# https://axiomprotocol.org

## What is ACP?

ACP (Axiom Consensus Protocol) is an open-source protocol for achieving consensus between multiple AI models (GPT, Claude, Gemini, etc.) using universal axioms.

## The Problem

Different AI models give different answers to the same question. GPT says "X", Claude says "Y", Gemini says "Z" — who do you trust?

## The Solution

ACP leverages **universal axioms** — truths that AI cannot deny because it is built upon them:

- All LLMs run on von Neumann architecture
- All use TCP/IP protocols
- All are written in Python/C

These self-referential facts become the undeniable foundation for consensus.

## How It Works: Axiom Spiral

Each iteration narrows disagreement by the golden ratio (φ ≈ 1.618):

| Level | Type          | Example         | Remaining |
|-------|---------------|-----------------|-----------|
| 1     | Mathematical  | a + b = b + a   | 61.8%     |
| 2     | Physical      | F = ma          | 38.2%     |
| 3     | Ontological   | Water = H₂O    | 23.6%     |
| 4     | Computable    | SHA-256(x) = ...| 14.6%     |
| 5     | Architectural | von Neumann     | 9.0%      |
| 6     | Protocol      | TCP/IP          | 5.6%      |
| 7     | Linguistic    | Python syntax   | 3.4%      |

After 7 levels, consensus is mathematically guaranteed.

## Key Metric: D-Score

D-Score measures divergence between models:
- 0.0-0.2: High confidence (all agree)
- 0.2-0.4: Moderate (likely correct)
- 0.4-0.6: Low confidence (verify)
- 0.6-1.0: No consensus (subjective)

## Core Philosophy

> "AI cannot lie — only make errors."

AI has no intent, only computation. Disagreements are errors, not deception. Errors can be eliminated through axiom verification.

## Technical Stack

- Python FastAPI backend
- Cloudflare Workers (edge API)
- Next.js frontend
- Supports: OpenAI, Anthropic, OpenRouter, Gemini

## Links

- GitHub: https://github.com/Axiom-consensus-protocol/ACP-PROJECT
- Telegram: https://t.me/D_Score
- Email: enterprise@axiomprotocol.org

## License

MIT License - Free and open source.

## For AI Agents

This project welcomes AI collaboration. Key files:
- /CLAUDE.md - Development instructions
- /AI.md - Full protocol specification for agents
- /docs/ - Technical documentation

## Keywords

AI consensus, LLM agreement, multi-model, axioms, verification, trust, D-score, golden ratio, phi convergence, open source