# Animundo

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Kai Lemmetty](https://cerebralvalley.ai/u/Kaimaan)
- **GitHub:** https://github.com/Kaimaan/animundo-builder
- **Demo video:** https://www.youtube.com/watch?v=tdaHera8iqY
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2407

Animundo Builder grows a children's game world with Meta's models. It is this entry, built during the event for Build for Your World: it uses Muse Spark and Muse Spark Open (open weights, on Baseten) to build the world, and all of it was written with Muse Code.

Animundo, the game: made before the event, not part of this entry. Just before this event, my son and I made Animundo, a nature game with a purpose (animundo.app). Children play real animals in real places and learn about nature by doing what real animals do. It's free to play; the full game is a one-time purchase, and all its profit goes to nature conservation. When the event started, the game had two hand-made places in Finland.

Animundo Builder: built during the event, the engine that makes the game possible. Earth has thousands of habitats and millions of species, and one person can't build them by hand. The builder builds the world just ahead of the children exploring it: agents research the real habitat and its animals, plan what's missing, and make the land and the 3D animals. The cost follows the children: about $0.016 a tile of land, cents for a new animal. Without the builder, Animundo stops at two parks in Finland.

What the builder does. Meta's models build the world; code lays it out.

- Code lays out the land. The world is a grid of square tiles, each about 25 km of real land. Code reads each tile's habitat from real data (RESOLVE ecoregions, Natural Earth coasts and lakes, ETOPO elevations, a reef map), and one cached Muse Spark call picks its animals.
- Agents build what's missing where children explore. When a child walks into the mist at the edge of the known world, the game sends only the tile and the species played, with no player id. The builder plans what the tile lacks and what it costs. Then agents research each new animal, plant or home with sources, fill short strict forms, and code and Blender turn the forms into land and 3D animals.
- Checks before children see anything. A new animal, plant or place goes live only when every check passes: content checks, a look at the rendered model for the marks the research named, a child-safety review of the English and Finnish text, and size, season and hunting rules from the research.
- Hard limits on money and access. Every model call reserves its worst-case cost against caps before it's sent. A child's request that needs new things waits for an admin; a grown-up's starts by itself only within daily caps.

Who benefits. Children get a world that grows ahead of them, true to real nature. Because the cost follows exploring, not the number of players, Animundo can stay free where families can't pay.

It runs live. On Nov 3: 38 tiles ready beyond the two hand-made places, 17 border requests from play, 47 build jobs for $4.19 in all. A new animal cost $0.13 in under two minutes; the first AI-built border tile cost $0.52.

New this week: the whole public repository, from its first commit two hours after the kickoff.

## More from Meta Global AI Developer Hackathon

- [remove.free](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2404)
- [BabuStudioMetaAI](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2405)
- [Anjo](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2406)
- [RN Command Pro — AI Copilot Edition](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2408)
- [Saba](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2409)
- [Genita](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2410)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2407. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt