@acdhaChris Adams @treyhunnerTrey Hunner 🐍 I would argue that types can be very "real" when defining a contract, but in a legacy codebase where you are adding them, the types *themselves* need tests. any modification to a type signature that is not validated against other typed code with tests of its own (and some heritage in real-life use) is an opportunity to just guess at the type, guess *wrong* and then start mangling the implementation to line up with an incorrect guess