In Haskell, it's the norm I think to do both: let the compiler infer the type and then add the inferred type to the code. This locks in compatibility.