The End of Apps As We Know Them

Lots of people trying to figure out what replaces the "grid of apps" approach to mobile. This article is a pretty good extrapolation of trends we're already seeing - first we had notifications. Then we had notifications you could interact with. Eventually, maybe, you'll get so many notifications you can't deal with all of them - and here PMs start salivating at the idea of being able to apply machine intelligence to the problem of what you need to look at next. (If I didn't invent the phrase "product managed to death" I don't want to know about it.)

What this approach under-emphasizes, I think, is user intent as an unpredictably evolving thing. We'll definitely get better at inferring your intent from your schedule, from your past behavior - but in a deep way user intent will always be impossible to model well. It's arguable how much this matters: most people have a daily routine, a set of boundaries they rarely go beyond. So maybe we'll be uncannily good at this most of the time. Maybe we'll even be able to model people's desire for novelty along different dimensions to keep their stream stimulating (although it's funny that we talk about stream optimization as a solved problem given how primitive efforts are today.)

But what's the UI for expressing intent above the level of a card in this model? Text or voice control, maybe, but that's just a modality. What kinds of intents - or preferences, or states - could be expressed that the system would be able to incorporate into its model? If I want something that isn't "find me X", are we back to a grid of app icons?