What overhead? More tokens for same result?
I haven't looked into it properly but, it looks like it might actually be the same amount of tokens? (Excepting stdlib boilerplate etc.)