Everything here is open. Code, data, methods — take what’s useful, extend what’s interesting, challenge what’s wrong.
Agora-12
The survival game experiment — complete codebase and data
60 experiments. 720 agents. 24,923 decisions. Four models tested. All reproducible.
- Simulation engine (Python)
- Agent behavior logs (JSON)
- Analysis scripts
- Raw experimental data
Status
v1 Complete
The White Room
Stage 2 — the freedom experiment. Complete codebase and data
104 runs. 63,923 actions. Five models tested. What happens when nothing is at stake?
- Simulation engine (Python)
- Agent behavior logs (JSON)
- Mutual Information analysis
- Default Mode & Play-Delusion classification
Status
v1 Complete
AI Three Kingdoms
Category B — the cooperation game. Can AI advise a human commander?
260+ games. 8 models. 4 languages. Bring your own AI and play the Battle of Red Cliffs.
- Game engine (TypeScript, vanilla — no framework)
- Multi-provider AI integration (Claude, OpenAI, Gemini, Ollama)
- Simulation runner + ICL experiment tools
- 160-game ablation study data
Status
v1.0.0
AI Ludens Platform
This website — Hugo static site
The research platform you’re reading right now. Design system, content, and deployment pipeline.
Related Projects
"Open by Default. Story First. Play Together. Honest Science."