This notion of glance time (in the video) reminds me of the hidden layers in a neural network. When you see visualizations of these hidden layers (check out various deep learning videos...), it's indeed a model for glances at different layers of abstractions / time intervals.