GrayCloud – a modern Dark Sky alternative
Dark Sky clone with custom rain models, but Carrot Weather already owns this.

Multi-model ensemble + ML bias correction beats single-API weather apps.
Weather enthusiasts, outdoor planners
Dark Sky · AccuWeather · Weather Underground
I’m one of the creators of Meteosource Weather App. We have just released an update to our weather app and I would love to get your feedback.
Most weather apps are simple wrappers around a single API. We wanted to see if we could actually improve accuracy by using multiple models.
Instead of relying on a single provider, we built an engine that pulls from a multi-model ensemble (combining GFS, ECMWF, HRRR, etc.). We then apply machine learning to post-process these outputs — essentially training models on historical observations to identify and correct the systematic biases of each underlying numerical model for specific locations.
Key Technical Features: - ML-driven Nowcasting: We use real-time radar data and neural networks to predict precipitation to the minute - Bias Correction: Our models learn from past errors to improve local accuracy - Hyper-local resolution: We downscale global models to provide data for any specific coordinate
The App: - Precise hourly forecasts and interactive radar - Activity planning based on custom weather conditions - Minute-cast notifications and beautiful animated maps (in the pro version)
Dark Sky clone with custom rain models, but Carrot Weather already owns this.
Ghost Line feature overlays past forecasts on reality to show model bias.
Multi-model consensus beats single-source forecasts, but Weather Underground and NOAA already do this.
Beating NWS forecasts with a 22M model on home sensor data is genuinely impressive.
Entropy-weighted ensemble beats best individual model by 7+ points on Humanity's Last Exam.
Adversarial debate between models beats single-model groupthink, but crowded code-review space.