Might as well use the zero-size struct{} instead of bool, given that you’re not currently using the bool values for anything: https://go2goplay.golang.org/p/9iegVQ2VQCr

Alternatively, use the bool values and rely on the zero value for missing elements: https://go2goplay.golang.org/p/E7yHQqAPseG

And variadic args to NewSet looks a bit nicer yet at the call site: https://go2goplay.golang.org/p/fTp3W9IwuVP

Is there a way to do this without losing range?

The proposal is, for better or worse, quite clear about not providing operator overloading or anything similar in the near future.

You could always offer a method that returns a slice of unique elements, which could then be ranged over.

I don't think there's nearly as much of an appetite for operator overloading as there is for an iterator protocol for range. That isn't to say that the latter wouldn't have a lot to argue over (internal vs. external, etc.) but I think the general usefulness of such a thing would be widely accepted by the community once generics are a settled matter, while I doubt that will ever happen for operator overloading.

I'm not sure that hiding the implementation is worth it here. Why not just make it public that it's actually map[T]struct{} underneath - then anyone can range on it and implement allocation optimisations, etc? I added some other methods too for the crack https://go2goplay.golang.org/p/EI1hYaSohnc