Built this over a weekend to experiment with AI-assisted coding.

Data comes directly from Germany's official energy registry (Marktstammdatenregister / MaStR) — 40,500 wind turbines with real GPS coordinates, capacity, manufacturer, hub height, and commissioning date.

Technical details: - The bulk export is a 2.8GB ZIP. Instead of downloading it all, I used HTTP range requests to read the ZIP central directory (42KB), locate EinheitenWind.xml, and download only that file (7.5MB compressed → 184MB XML)

  - Rendered with deck.gl + MapLibre (CartoDB dark tiles, no API key needed)

The EEG layer is what surprised me most: 9,247 turbines have already lost Germany's 20-year wind energy subsidy. Another 3,094 expire by 2028 — those are the repowering candidates.

Happy to answer questions about the data pipeline or rendering approach.