These findings point toward a question that deserves serious study. When a machine returns an answer that is polished and confident, it removes not only the labour but the experience of struggle — the uncertainty, hesitation and difficulty we rely on to sense whether we are actually learning. We interpret the ease of achieving an outcome as evidence of competence, but it may be nothing of the kind. I call this a fluency hijack: the suppression of the very signal that would warn us that learning is not happening, leaving a gap between the competence we feel and the competence we hold. What the science of learning has long understood is that some difficulty is not an obstacle to be overcome but the work itself. Psychologists call these desirable difficulties. It is the effort and productive struggle that turns out to be the very thing that imprints a durable skill into the mind. A fluent answer from a large language model removes the desirable difficulties worth experiencing.

The human judgment developed through struggle allows us to know when to let the machine carry the load and when to carry it ourselves. Judgment of that kind can't be downloaded or upgraded. It is built the same way every other expertise is built - through the effort we are now tempted at every turn to skip. The risk, if we skip it, is a slow skill atrophy we won't notice until we need it, and this can happen while every business dashboard looks healthy. We must not run our companies, our classrooms, or our own minds on autopilot. Autopilot is a marvel when someone capable is watching the instruments, and a potential catastrophe when no one is. The same is true of "intelligent" machines. To adopt AI is to stay awake at the controls and continue to ask what we are gaining, what we are surrendering, and whether the people in our companies are still growing or simply producing. What gives me hope is that none of this asks us to choose between the human and the machine, rather to use these tools deliberately, embrace the difficulties that make us capable, and treat judgment as the thing most worth cultivating. We have always been a place that learns the most by doing hard things. Our task, then, is to carry that same mindset into this new moment — to let the machine do what it does well, and to continue to struggle with the human work that makes us who we are.