Galaxy Brain Why Is It So Hard to Make a Good Weather App?
22 snips
Mar 13, 2026 Adam Grossman, physicist and co-creator of Dark Sky turned Acme Weather founder, explains how phone context reshaped forecasts. He recounts building minute-by-minute radar predictions, cleaning messy station data, and why owning the service matters. He also talks about designing UIs that show uncertainty and how ML and generative tools could change weather communication.
AI Snips
Chapters
Transcript
Episode notes
Road Trip Rain Sparked Dark Sky
- Adam Grossman built Dark Sky after a torrential rest-area downpour during a 2010 road trip that revealed existing apps' uselessness.
- He prototyped minute-by-minute radar prediction to tell users when rain would stop, which became Dark Sky's original focus.
Forecasts Start With Observations Not Guesses
- Forecasting begins with massive observational data from satellites, balloons, stations, and buoys to define the atmosphere's initial state.
- Those observations feed numerical weather prediction models that run physics simulations on supercomputers to produce forecasts.
Data Sanitation Caused Embarrassing Forecast Bugs
- Early Dark Sky suffered wild data errors like stations reporting implausible temperatures, producing forecasts off by hundreds of degrees.
- Grossman says most forecasting work is sanitizing and munging station data to avoid such glaring failures.