How did input on HoloLens get to where it is now, what have we learned about it, and where are we going next?