Ten years ago I added a reminder to my calendar to check on how this prediction was panning out:

In a decade’s time computing seems likely to take the form of AR interfaces mediated by AI, using gestures and speech for inputs and the whole world as its display.

Speech input has come on, and LLMs have happened. AR appears to have totally failed, but maybe it just needs another 10 years.

From: Imperial ambitions, The Economist, 9 April 2016.