It's comparing static typing as exists in Haskell or OCaml to static typing as exists in C++ or Typescript. To dynamically type a program in the first group, you would have to manually make a giant type that includes everything and use that everywhere. In the second, you can program dynamically by turning the safety off.
Are you sure you meant C++ in the second group? In Typescript you can program dynamically, perhaps. In C++, you'd have to cast everything brutally, which I don't think most would call programming dynamically, or do what you propose for OCaml.
Casting at all really solidly puts it in that group. The idea that there’s a program and a type system and you tell the type system when it’s diverged from the program doesn’t really work in the other group of languages.
(Complete rewrite) I think this may be true in theory but not a practical way to think about c++, as the language fights you if you try to use in that way.
