Surprising a model
Back when my kids were little and I was working at Google I was dogfooding an internal Maps or Google Now (remember that?) experiment on my Android phone that nudged you to get going for an event that it had inferred you needed to get to—even if it was not explicitly an event on your calendar!
For me, this was a notification at 5:30 pm on weekdays to get moving to pick up kids from daycare before it closed at 6 pm. It understood this about my life with very high confidence after maybe a week of silent observation.
I separately looked up my location history and noticed how much it was like clockwork. A perfect triangle between home, daycare and the office. Clockwise in the morning, counter-clockwise in the evening, in time windows varying by a few minutes.
I remember thinking to myself: the bulk of my life can be captured in—and predicted by—a protobuf that is probably under a kilobyte. On a personal level this gave me a lot of food for thought.
But on a philosophical level it drove home another point: one way to look at agency or free will is the ability to surprise a predictive model.