Wind God — Aeolus

About

Your spot doesn't exist in any weather model. So we build it. Global weather models compute grid cells of 2 kilometers and more — your home spot is literally invisible in them: no lake, no shoreline, no venturi effect, no thermal edge. Yet that's exactly where your wind is born. Aeolus closes this gap with an approach that until now existed only in research. WHAT YOU GET • Spots today: Reschensee, Walchensee, Ammersee, Kochelsee, Lago di Garda, Egypt El Gouna — plus 7 in testing: Starnberger See, Alpsee, Chiemsee, Forggensee, Lake Constance (Friedrichshafen, Güttingen, Konstanz). The library keeps growing. • Kiteable probability for every hour — including your personal threshold (foil? big kite? set your limit). • Wind alerts that are truly yours: set them per spot — wind direction, strength, minimum duration, and how confident the forecast has to be. • Regime intelligence: foehn and thermals are different physics — our models treat them that way. • Instant updates with every weather-model run — precomputed, the answer arrives immediately. • Free to use throughout 2026. After that, Aeolus will likely cost a small fee — covering compute costs and the development of new spots. Aeolus is not another weather widget. It's a research program growing into your pocket — built by simulation specialists, tested in the field, computed with the physics of your spot. THE AEOLUS PROGRAM: A DIGITAL TWIN FOR EVERY SPOT Every spot gets a complete digital twin: real topography on a 50-meter grid, terrain roughness, water surfaces with their temperature, the energy balance of every single slope. On top of that, a 3D multiphysics flow simulation (CFD + turbulence modeling) computes the wind field physically: moist air, solar radiation, thermals, foehn, lake breeze — alpine lake, foehn corridor or coastal lagoon, every spot's physics under the same laws of nature. The first chapter is a showpiece in the Bavarian Alpine foothills: Kochelsee and Walchensee — two lakes, 200 meters of elevation apart, a mountain ridge in between — in a single simulation domain. One model, both spots, including their interaction: whether foehn- or thermally driven, both run in the same computation. The physics is fascinating. But what matters to you is: is the drive to the lake worth it? THE NEURAL SURROGATE The full simulation costs hundreds to thousands of CPU hours per spot. That's why the heart of the program is a 3D neural surrogate model: a neural network trained on very many complete 3D simulations per spot. It doesn't memorize weather statistics — it learns to solve the physics itself: the result of the simulation in seconds on a single GPU, a speedup factor of 10,000 to 100,000. The same technology class NVIDIA's FourCastNet used to accelerate global weather forecasting from supercomputer hours to GPU seconds — except FourCastNet computes the globe at 25 kilometers. Aeolus brings this class to your spot: 50 meters. And because the training data comes from our own simulation, a new spot doesn't need ten years of measurement history — just geodata. 100,000× faster than the physics it learns from. Physics accuracy, app speed, scaling by blueprint. The wind has always been there. Now you can see it coming.