I'm not sure why you're being downvoted. This technique is not entirely unlike using a phantom type, which is a common way of encoding such constraints into th

I'm not sure why you're being downvoted. This technique is not entirely unlike using a phantom type, which is a common way of encoding such constraints into the type system. I'll admit it's not ideal; in fact, it's often downright cumbersome. In some cases, it's not possible to encode desired information into a particular type system. But in this case, it seems valid to me. There is no way to construct a value of IntegerPowerOfTwo that does not represent a power of two, and therefore it fulfills GP's example requirement.