ClimApp Pance
Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

ClimApp Lite tells people in the Pance river basin (Cali, Colombia) whether heavy rain (≥20 mm) is likely tomorrow, and lets them report what they see on the river, verified against real data before anyone trusts it. A calibrated random forest trained on local rain-gauge stations gives a daily probability, compared with the seasonal normal and a 3-day accumulation percentile. Muse Spark 1.3 explains the forecast in plain Spanish (every number checked by code), and runs a verification gate for citizen reports: it structures the free-text report, a second Muse call judges it against rain gauges, river level and NASA IMERG satellite rainfall, and deterministic code computes the confidence. Exaggerated or unsupported reports go to human review instead of being published. Evaluated on 50 simulated reports with data-derived labels: 80% correct within 3 attempts, 0 exaggerated or false reports verified, US$0.021 per report. Open source, with limitations documented.