N.B. The incoming paper did say U here, but I think this is editorial because it's inconsistent with the rest of the std::polymoprhic spec. The constructors taking in_place_type_t<U> to specify the type all require is_same_v<remove_cvref_t<U>, U> to be true, so these two constructors are inconsistent. It's unclear what it would even mean to store e.g. const X& as the owned object. So remove_cvref_t<U> is the only reasonable interpretation, and that's what UU is.