According to CoinMeta, Google has published the latest research on WeatherNext in Nature. WeatherNext is a weather forecasting model that can predict the path, intensity, and wind patterns of tropical cyclones. Tests have shown that, at the same level of accuracy, it can forecast about 24 hours earlier on average. For example, forecasts made 3 days in advance have reached the same level of accuracy as those made 2 days in advance by other models. This improvement is mainly reflected in the prediction of hurricane intensity. Previous AI models could determine the path of hurricanes but struggled to accurately predict their final intensity. For disaster warnings, knowing when a hurricane will make landfall is not enough; it is also necessary to predict its intensity as early as possible. For the 2025 hurricane season, the National Hurricane Center of the United States has started using WeatherNext for real-time forecasting references. Five days before hurricane Melissa made landfall in Jamaica, WeatherNext predicted with 80% certainty that it would make landfall at category five intensity, and this probability approached 100% when the forecast was made 3 days in advance. This is the first time that the National Hurricane Center of the United States has predicted that a hurricane, which was only at category one at the time, would eventually intensify to category five.