The weeks before the map

A city asks, “Which neighborhoods should we check first after the next heat wave?” The hard part is not drawing a map. Someone must find data that match the same places and dates, clean it, choose useful signals, and test whether the answer works outside the examples used to build it.

Google’s Planetary Prediction Engine is designed for this bottleneck. It automates much of the path from a location-aware question to a candidate model. That does not make the answer certain. It makes the first serious experiment cheaper.

The question for the rest of this tutorial is not “Can AI know the planet?” It is “How can faster experiments help a real team decide what to verify first?”