# Catan Bias

- **Event:** [WeaveHacks 3: Self-Improving Agents Hackathon with Weights & Biases](https://cerebralvalley.ai/e/weave-hacks-3-self-improving-agents-hackathon-with-weights-and-biases-7014fe80)
- **When:** Jan 31 at 9:00 AM – Feb 1 at 5:00 PM (PST)
- **Where:** 400 California St, San Francisco, CA
- **Team:** [Petros Hong](https://cerebralvalley.ai/u/petroshong)
- **GitHub:** https://github.com/petroshong/catanlearn
- **Demo video:** https://youtu.be/n1E8BfAfQkM?si=5yszReCDLYbkWF1j
- **Gallery:** https://cerebralvalley.ai/e/weave-hacks-3-self-improving-agents-hackathon-with-weights-and-biases-7014fe80/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/weave-hacks-3-self-improving-agents-hackathon-with-weights-and-biases-7014fe80/hackathon/gallery/28

Full browser Catan game built with React, TypeScript, Tailwind, and Supabase. Includes hex board, roads, settlements, cities, dev cards, trading, robber, longest road, and victory points. AI uses heuristic placement/build logic with optional LLMs (Gemini, Anthropic, MiniMax, OpenAI) for chat and trades. Python self-play (python/train.py) tunes weights and logs to wandb, exporting wandb-summary.json, which the frontend loads so LLMs and browser AI share the same training context. Marimo provides a dashboard to run games and track win rate and rounds.

---

Markdown version of https://cerebralvalley.ai/e/weave-hacks-3-self-improving-agents-hackathon-with-weights-and-biases-7014fe80/hackathon/gallery/28. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt