Leila is twenty years old, studying history, and not cheating. She uses AI to explain difficult passages, suggest essay structures, and give feedback on drafts. She is engaged with every response. She is making good decisions with what the AI gives her. And something is happening to her thinking that she doesn't yet have the vocabulary to describe. By the third month of term, she finds it harder to sit with a difficult paragraph without reaching for her laptop. The discomfort of not immediately understanding something — a discomfort she used to work through — has come to feel less like productive tension and more like a problem to be solved.

This post is about what is being quietly replaced. It draws on a striking MIT preprint that measured brain activity in students writing with and without AI — and found what the researchers called "cognitive debt." It draws on fifteen years of learning science from Justin Sung, whose new video (published this week) makes the mechanisms visible with unusual clarity. And it connects to something I have been thinking about since the health monitoring essay: the tool works, and something is quietly replaced by it. The question is not whether students are submitting AI work. It is whether they are losing the capacity to think without it.