Why don't you just run ty on the ty codebase and let it tell you which errors are correct and which are erroneous?