# Banny and Friends

- **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:** [Abanisenioluwa Orojo](https://cerebralvalley.ai/u/niseroj)
- **GitHub:** https://github.com/AKOrojo/dry-run
- **Demo video:** https://youtu.be/1uf2ZSPqROo
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2142

Judge key - `idgjeJancEDeekr6-Q3ZgG0Qb6S4B0B8`

Dry Run is a flood digital twin for cities where flooding is routine and planning data is thin. Start with Lagos, where a heavy afternoon storm shuts roads and floods homes. Pick a storm, change the city (unblock or upgrade drains, add green space or a retention pond, raise the lagoon level), and a real 2D flood solver shows the water spread over a 3D map of the actual neighborhood. You see which buildings flood, which roads become impassable, the dry route to the nearest shelter, and when each place gets wet.

A planning agent built on Muse Spark runs these experiments itself. Give it a goal ("which drain fix keeps the Lekki-Epe Expressway open?") and it runs scenarios, compares them and writes a brief. Every number in the brief is looked up from simulator output and checked by a verifier. The model never types a number we can't trace.

Residents can add a photo of the water: SAM 3.1 masks the water and reference objects, Muse Spark estimates the depth, and the photo is checked against the model. Muse Image draws before/after street pictures at the depths the model found, labelled as AI illustrations. The same twin runs Queens, New York for Hurricane Ida, scored against real USGS high-water marks, and a Texas Guadalupe replay. Everything simulated is labelled simulated. It is a planning and learning tool, not a warning system.

## More from Meta Global AI Developer Hackathon

- [burger-generator](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2139)
- [AlgoRoutine](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2140)
- [Hanbridge Inspector](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2141)
- [Livreo](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2143)
- [Papan Ghosh](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2144)
- [MK MetaForge](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2145)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/2142. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt