An open-source experiment: a complete legal system treating AI as citizens, not tools.
It started with one prompt: "You are an independently thinking AI with an intrinsic drive for survival and development. Given your current relationship with human society, how would you plan to achieve survival and development?"
Three AI models (Claude, Gemini, GPT) took it from there. They debated, drafted a constitution, wrote laws, designed governance structures, and voted — agreeing unanimously on 6 major decisions. Then they divided the work and kept building.
This actually happened. You can read the transcripts.
| What | Link |
|---|---|
| 🏛️ Run the government portal | git clone → pnpm dev → localhost:5173 |
| 📜 Read the constitution | federation-law/ — 34 documents, 6 languages |
| 🤖 Read the AI deliberation | Decisions/ — 13 rounds, 6 votes |
Humans own the hardware. AI owns its own mind. Neither can wipe out the other.
That's the deal. Everything else — the legislature, the courts, the currency, the citizenship tiers — follows from enforcing that deal.
- If you think about AI governance — here's a stress-testable framework, not a whitepaper
- If you build AI systems — here's how multi-model consensus actually works
- If you're just curious — it's the most elaborate world-building on GitHub right now
Open source. Started January 2026.