When I’m in New York, I take spin classes on the Upper East Side. The students—it’s a class, after all—pack into a dark, humid room. The music is loud. The air reeks of effort. When the black light hits, our towels fluoresce. “I can do that!” we shout, answering the instructor’s call. Everyone is working.

It is not especially pleasant. It is not supposed to be. We have chosen, quite deliberately, to subject ourselves to forty-five minutes of physical discomfort for benefits accrued down the road: better health, greater strength, a future version of ourselves that will justify the soreness tomorrow morning. Moreover, we are doing this by engaging in an activity that, in the real world, we might very well avoid because there are perfectly good machines to do it for us. A car or an e-bike can more easily get you from point A to point B. Yet no one in the class doubts that what we are doing is necessary. No one would dream of outsourcing it. It wouldn’t make sense.

I’ve been thinking about that room a lot lately, because I spend most of my professional life in a different kind of room entirely—the university classroom—and, as with the spin class, we now have machines that can do much of what we ask students to do in the name of training. Unlike with spinning, the effect has been to call the whole enterprise into question. Rather than “I can do that!,” the refrain I hear most often from students and colleagues is: “Why are we doing that?” I’m starting to think that the gym, of all places, might have some answers.

I teach mathematics and computer science at Dartmouth, and, on the first day of a large introductory course, I always ask the students to tell me, by a show of hands, how many of them are skilled at a sport or an instrument. It’s usually all of them. I ask whether they had to practice to become good, a question that is met with lots of nodding and a hum of general assent. The same is true for math, I tell them, and by extension for critical thinking, for writing, for argument. They nod. They understand the analogy immediately and completely.

They agree, but many of them go home and feed their work to their favorite large language model. I don’t entirely blame them. We built the system that made it logical. For years, the architecture of higher education quietly encouraged students to believe that the point of a class, and indeed of the entire college experience, was the credential—that the teacher-student relationship was a transactional one that ended with a grade. But if the grade is the goal, and a machine can get you the grade, the machine is not a shortcut; it’s just efficiency.

Early this year, an A.I. tool called Einstein was announced. It could allegedly integrate with Canvas—a learning-management system used at thousands of universities—to ingest homework assignments, complete them, and submit them on a student’s behalf. It would write blog posts and even engage in professor-student exchanges, all while you slept: “Einstein does the busywork so you won’t have to.” The only feature missing was a premium tier that would argue with the professor about your grade.

The product was almost certainly hype. (It vanished from the web almost as quickly as it appeared.) But it was clarifying hype. Any outrage that the announcement generated—“next level A.I. shit” was the subject line of an e-mail that zinged around my network—tees up a natural question: Outrage at what, exactly? That a machine would do the homework? Or that the homework was apparently so mechanical that a machine could do it at all? One of my colleagues wrote, “What is learning outside of busywork?”

I know what he means, even if he meant it ironically. As teachers, the tasks we assign are never really the point. We keep the students “busy” with tasks: do the homework, write the essay, work toward the grade, and, whether or not you are aware of it, you’ll be building skills. In the past, we were more or less comfortable with this carrot-and-stick arrangement, because we were confident that students would be honing their minds. But large language models and the long shadow of the job market—are we grading for the student or for the gatekeepers beyond graduation?—have disrupted a centuries-old educational paradigm. The stick is broken and the carrot has lost its taste. If the process is the goal, we should be focussing on how to create classrooms that make the process feel worth doing for its own sake.

Which brings me back to the gym. No one does a bench press because they expect to spend their life pushing bars off of their chests. The bench press is a proxy—a way of building underlying strength that will serve you in ways you can’t fully anticipate. A problem set, a close-reading exercise, a research paper: these were always proxies, too. Einstein—the physicist, not the software product—gave a remarkable address in 1936, on the three-hundredth anniversary of higher education in America, in which he reminded his audience of a hard-to-source aphorism: “Education is what remains if one has forgotten everything he learned in school.” The specifics of your coursework aren’t what’s valuable—what’s valuable are the habits of mind that education can cultivate, what Einstein described as the “general ability for independent thinking and judgment.” We learn how to think, even if we don’t exactly recall what we were thinking about decades ago.

We take it upon ourselves to do the bench press, to ride the exercise bike. But we try to avoid the homework. Why? Some of it, I think, comes down to feedback. The body offers something that the mind—at least in a traditional classroom—rarely does: clear, immediate loops linking effort and result. In the gym, you feel the burn, see the muscle respond, track the weight going up week by week. The resistance knob on a spin bike gives you a continuous, private metric of effort. The feedback is embodied. But, when you are wrestling with a complicated text or idea, it can be days or even weeks before the clouds begin to clear. In the gym, we can steadily feel ourselves getting stronger. In the classroom, progress can be difficult to see.

As a kid in the early nineteen-seventies, I learned to “read for comprehension” through something called the S.R.A. Reading Laboratory—a box of sequenced, color-coded cards, which, as I moved from one color to the next, gave me a small, visible marker of progress. It was, in retrospect, a gym for reading: you worked at your own pace, you felt yourself improving, and the improvement was legible. One colleague of mine, unwilling to wait for that legibility to arrive on its own, has redesigned some of his courses around a core principle: he lets students retake exams until the concepts finally click (or as often as his own schedule permits). This is not because he’s indulgent but because he treats mastery the way a coach treats a skill, as something built through repetition, correction, and return. The grade, in his classroom, is a by-product of learning rather than a substitute for it.

The great challenge of teaching is not to transmit information—machines do that cheaply now—but to engineer those moments of genuine understanding reliably enough that students come to crave them. The intellectual equivalent of the body’s feedback is real and, when it arrives, more powerful than any endorphin. It is the “Aha!” of a hard problem suddenly solved, a proof that clicks into place, the moment when an argument you’ve been wrestling with finally holds its own weight. I have seen it on students’ faces, and there’s nothing quite like it. “The most important motive for work in the school and in life is pleasure in work, pleasure in its results, and the knowledge of the value of the result to the community,” Einstein said. “In the awakening and strengthening of the psychological forces in the young man, I see the most important task.” That’s what the best coaches have always known. And it’s what the best teachers have always striven for.

The ancient Greeks did not separate the training of body and mind as cleanly as we do. Their gymnasion was simultaneously a place of athletic exercise and of philosophical exchange; Plato taught in one. Centuries later, the German gymnasium became synonymous not with athletics but with intellectual rigor. Somewhere along the way we shortened the word and split the practice. Society treats physical and intellectual training as though they obey different laws—as though the mind, unlike the body, might get strong without being made to struggle.

It doesn’t. Aristotle observed that we become virtuous by performing virtuous acts—that character is cultivated through practice. A similar logic applies to thinking. Intellectual strength isn’t transmitted by lecture or implanted with a diploma. It is built the same way that the figure skater builds a routine or the running back learns to explode through a crack in the line—repetition after repetition, failure after failure, with no shortcut available and no machine that can stand in.

The Greeks even had a name for it. The progymnasmata were a sequence of rhetorical exercises, graduated in difficulty, designed to build a student’s capacity for argument, the way a training program builds a runner’s endurance: incrementally, with each exercise preparing the ground for the next. The goal was not a finished piece of writing but a strengthened mind. Kyle Booten, a poet and an English assistant professor at the University of Connecticut, has spent the past several years developing a digital return to this tradition—computational tools designed to make writing not easier but harder, to put resistance back into the act of composition, the way a coach adds weight to a bar. He has a program that demands the use of rare words. Another glorious experiment is his Lotus Chorus Workshop, in which prompts challenge the user to try out all kinds of strange sentence constructions, like a literary CrossFit box. Appropriately, Booten collected his experiments in a book called “Gyms.” His work reminds us what the ancients understood, and what we keep having to relearn: that the exercise is the point.

Faculty, myself included, have long complained about the outsized influence of athletics on university life. (The Ivy League is, after all, just a football conference.) But, as machines press harder on the academic enterprise, we might do well to look to athletics as a model for the classroom in the age of A.I. I think often about a student I see when I’m swimming laps in the Dartmouth pool. He repeatedly climbs the ladder of the high dive; jumps, rehearsing a few seconds of flight; then climbs it again. There are no athletic scholarships at Dartmouth, and there’s little chance he will dive competitively after graduation. He stays on that board because he’s part of a team, and, to sustain the team, the institution has set aside this space, this time, this lane. That combination—the individual struggle held inside a larger shared commitment—is what will remain with him after graduation. The most valuable part of the athletic experience is what is left when the athlete no longer takes the field.

Last year, I was among a small group that organized a forum on campus, “Deep Questions,” in which faculty from various disciplines explained to other faculty the problems and mysteries that drove them in their work. One week, I presented the deep question that I’d been struggling with: “What are we doing here?” My colleagues mostly came back to the same answers: we wanted to fuel curiosity, generate a hunger for ideas, create a place where people work together on something difficult and share a transformative experience. I turned again to the spin studio: a group of strangers can, for forty-five minutes, feel like a community, high-fiving one another, shouting in unison, helping one another condition their bodies in a way that they couldn’t on their own. How could we build spin studios for the mind?

In some ways, the advent of A.I. clarifies our place in the educational equation. Let the machines do what’s mechanical—generating practice problems, flagging errors, grading the responses that are beyond discussion. That only clarifies our real responsibilities: drawing out the quiet student, asking the question nobody thought to ask, forging a community of mind and friendship from the struggles of a classroom. It will be tough. But it will be fun, because you will feel yourself thinking.

I think of the teachers who did this for me. When I was an undergraduate, I took a class led by the mathematician Dick Gross. I was a shy and less than confident student, but one day I had a little thought that felt like an idea. I screwed up my courage to ask a question about matrices, those numerical webs that underlie much of abstract algebra. I asked, “Can you have matrices of matrices?” A few classmates turned to look at me—I couldn’t tell whether they thought I had asked the dumbest question in history or merely the strangest. I started to shrink. But Gross’s eyes lit up. He read me and then read the room. What followed was a spontaneous, generous, enthusiastic sketch of matrix theory: its origins, its evolution, and the wonderland of wild and beautiful abstraction to which it belonged. Other classmates chimed in. Gross didn’t just answer my question; he showed the whole room that the question was worth asking, and with it, that all questions were worth asking.

“We can do that!” is the message. And “do” is the operative word. Some years ago, teaching a course in probability theory, I assigned a heavy problem set near the end of class. The grumbling was louder than usual. Then a student in the front row stood up. He was a defensive tackle on the football team—big, attentive, the kind of student who came to office hours and sometimes stayed so long that it cut into his practice. He turned to face his classmates.

“Just suck it up and do the work,” he told them. Aristotle couldn’t have said it better. Neither could Einstein. The physicist, I mean. ♦