@brentsimmons (a second attempt, this time maybe with words that make some sort of sense)

I agree and think this was inevitable, since the interaction model is limited and so the software has to bend over backwards for even the simplest thing. As compared to the indirect model, where you can build things together and more easily make both simple and complicated software. For this reason I think the current form of touch-and-voice is a transitional step when compared to the rock-steady finality of text-and-cursor.