Banny and Friends

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.