Imagine being aware of the exact location of a hurricane’s landfall, the intensity of its winds, and the amount of rainfall it will produce many days in advance.

Thanks to a group of 20-something Stanford graduates who are gaining ground in the cutthroat industry of artificial intelligence-powered weather forecasting, that day might already be here.

The AI weather prediction champion, Google’s DeepMind, was surpassed by startup WindBorne Systems on Wednesday in significant benchmarks established by official weather models in the United States and Europe.

In an exclusive interview with Semafor, the company’s co-founders, who are supported by Khosla Ventures and Footwork, revealed how they collected detailed data using ChatGPT’s AI algorithms using an in-flight fleet of about 100 low-cost, manually constructed weather balloons made of plastic that they bought from a restaurant supply company.

The Semafor analysis claims that WindBorne’s WeatherMesh system capitalizes on two technological trends: the sharp drop in the price and size of computer hardware and wireless equipment, as well as the quick development of AI algorithms.

WindBorne’s weather balloons use artificial intelligence (AI) to accurately manage their routes so they may orbit the planet for weeks at a time, for around the same cost as a low-priced cell phone. With plans to grow its fleet of weather balloons 100 times to 10,000, WindBorne claims to already have the largest constellation in the world. This gives the modest business the same level of insight into Earth’s weather processes as well-funded government institutions.