Bidirectional is a nice, pragmatic compromise, but it is a reaction to the difficulty with taking full type inference beyond let-polymorphism. We wouldn't be using bidirectional, if Damas-Hindley-Milner would scale.
I dunno, I do prefer having type directed editing (eg. auto case splitting, and autocomplete), so full type inference is less of a holy grail to me.
Damas-Hindley-Milner inference is also hyper-optimised to a specific point on the design space, and does an excellent job there, but (to echo Conor McBride) I would love to see less conflation of the ideas of execution phase, parametericity, and implicitness going forward.