The Scene
My son Bennett is six years old and completing kindergarten this spring. He is learning to read. He is learning to count. He is, by every measure I can observe, becoming a person who thinks.
One evening not long ago he came home from school with a school-issued iPad tucked under his arm. This was unremarkable — the device had been part of his classroom routine for months. What was remarkable was what happened when he opened it at the kitchen table, found an application, and slid it toward me with the casual pride of a child sharing something important.
"This is math," he said.
I watched. On the screen, a cartoon character moved through a landscape collecting coins. Numbers appeared in brief flashes. Tapping the right answer triggered a small explosion of color and sound. Tapping the wrong one prompted a gentle chime and a second chance. There was no pencil. There was no counting on fingers. There was no pause long enough for anything to be hard. The game rewarded speed and compliance, and it was engineered — with obvious sophistication — to feel exactly like what my son believed it to be.
Math.
I sat with that word for a while. Not because my son was wrong to call it math — the school had handed it to him as math, the app presented itself as math, and at five years old you name things by what the adults around you call them. I sat with it because of what the word was doing in that sentence: covering for something. Describing a shape where the substance had been quietly removed.
What Bennett was doing on that iPad was not math. It was the performance of math. It was math with the thinking taken out.
I want to be precise about this, because the distinction matters more than it might initially appear. I am not a nostalgist making the familiar complaint that screens have replaced wholesome analog childhoods. I am not arguing that technology has no place in education. What I am arguing is something narrower and, I think, more troubling: that we have allowed a particular kind of substitution to happen in our schools and in our homes, and we have called it progress because the thing being substituted looks and feels enough like the original that most of us haven't stopped to check.
This blog exists because that substitution is not confined to kindergarten classrooms. It is happening across the full span of human cognitive life, and it is accelerating. The five-year-old tapping at a screen to collect coins is the early version of the adult who asks an AI to draft the email, summarize the document, form the argument, finish the thought. The mechanism is the same. The friction is removed. The reward arrives. The capacity that would have been built by doing the hard thing goes unbuilt.
We have a name for what happens to a muscle that goes unused.
Atrophy.
That is what this is. And it starts, as I discovered on an otherwise unremarkable Tuesday evening, earlier than most of us want to believe.
The Assumption We Never Examined
Nobody decided that six-year-olds should learn mathematics on iPads. That is the first thing to understand. There was no debate, no controlled study presented to parents at a school board meeting, no moment at which someone in authority stood up and said: we have weighed the evidence, and screens are the right medium for building numeracy in young children. It did not happen that way. It happened the way most consequential changes in education happen — incrementally, commercially, and beneath the threshold of serious scrutiny.
The EdTech industry — educational technology, in the full, earnest rendering of the term — is a market worth well over a hundred billion dollars globally, and it sells to some of the most motivated buyers in the world: administrators who need to show progress, teachers who are overextended, and parents who have been told, for the better part of two decades, that their children will be left behind if they are not digitally fluent from the earliest possible age. Into this climate of institutional anxiety, the iPad arrived not as a question but as an answer. It looked like the future. It felt like initiative. And it came bearing a word that effectively ended the conversation before it started.
Educational.
That word is doing extraordinary work in our culture right now, and almost none of it is being examined. Call an application educational, and it acquires a kind of moral immunity. Parents who would scrutinize the sugar content of a breakfast cereal do not scrutinize the cognitive content of a kindergarten app. Administrators who would demand outcome data before adopting a new reading curriculum accept "engagement metrics" — time on task, completion rates, stars earned — as proxies for learning. The label launders the product. It converts a commercial transaction into an act of educational virtue, and it does so without requiring anyone to define what, precisely, is being learned.
Ask that question in the wrong room and you will be looked at with mild suspicion, as though you have raised an objection to clean water. Of course it's educational. It says so right there. The children enjoy it. The teachers have time for other things. What exactly is the problem?
The problem is that enjoyment and learning are not the same thing, and we have built an entire infrastructure on the assumption that they are — or that the gap between them doesn't matter, or that a child who is engaged is a child who is growing. These assumptions have the texture of common sense. They are not. They are sales propositions that have been repeated so many times, by so many well-meaning people, that they have quietly achieved the status of fact.
Consider what it takes for a school to adopt a piece of educational software. A vendor makes a presentation. The interface is clean, the animations are appealing, the data dashboard is impressive. Someone mentions that another district is already using it. A pilot is proposed, perhaps run over a semester in one or two classrooms. The pilot "succeeds" — meaning the children used the software regularly and the teachers found it manageable. The product is adopted. It enters the classroom as an established tool, and within a year it is simply part of how things are done. No one asks whether the children who used it learned to read or count better than the children who did not. That question is rarely built into the pilot. It is rarely asked afterward. The machine is already running.
This is not a conspiracy. The administrators involved are not cynical people. The teachers are doing their best under conditions that have become genuinely difficult. Many of the people who build educational software believe sincerely in what they are making. The problem is structural: the incentives at every stage of this process reward adoption, not outcomes. A school that adopts new technology signals modernity and ambition. A school that questions whether the technology works signals resistance and complacency. The bias is baked in, and it operates long before any individual makes a conscious choice.
What gets lost in this structure is the child. Not as a rhetorical abstraction — as an actual six-year-old sitting at a kitchen table, iPad in hand, telling his father that what he is doing is math. The child who has been given a game calibrated to feel productive, whose every correct tap produces a small dopamine reward, who is accumulating screen time and stars and a sense of forward motion while the slow, resistant, unglamorous work of building real numerical intuition goes unscheduled and undone.
We did not choose this for our children. We allowed it to be chosen for us, by an industry that had every reason to make the choice feel inevitable. And we allowed it because we never stopped to interrogate the one word that made it all seem not just acceptable but forward-thinking.
We never asked what "educational" actually means.
What Learning Actually Requires
So let us ask it. What does "educational" mean? What is actually happening inside a child's mind when genuine learning takes place — and what is not happening when it doesn't?
Start with mathematics, because that is where we began. When a young child learns to count, something specific and physical is occurring. The child is not merely memorizing a sequence of sounds. She is building a mental model of quantity — a felt sense of what "five" is, what "three" is, and what it means to combine them. Researchers in cognitive development call this number sense, and it is not a trivial precursor to arithmetic. It is arithmetic's foundation. Without it, a child can learn to produce correct answers without understanding what those answers represent — which is a precise description of what most mathematical games on a screen are training children to do.
The physical dimension of this early learning is not incidental. When a child counts on her fingers, she is not using a crutch that will need to be unlearned later. She is doing something cognitively sophisticated: linking an abstract symbol to a concrete, embodied experience. The finger is a bridge between the number as a word and the number as a quantity. Neuroscientific research has shown that the region of the brain responsible for finger representation is closely linked to numerical processing — so closely that children with stronger finger awareness in early childhood consistently demonstrate stronger mathematical ability in later years. The finger is not a shortcut around thinking. The finger is part of how thinking about numbers is built.
A gamified math app eliminates the finger. It eliminates the pause, the recount, the moment of uncertainty in which a child holds a quantity in her mind and works to manipulate it. It replaces that moment with a tap. The correct answer is usually derivable by elimination and pattern recognition — skills that are real, but are not the skills the lesson claims to be building. The child learns to play the game. The game is not the same as the subject.
Reading development follows the same logic, and the evidence here is if anything more emphatic. The past two decades of cognitive science have produced a robust and remarkably consistent body of research on how children learn to read — a body of research so well-established that it now goes by a proper name: the science of reading. Its central finding is not complicated. Reading is not a natural act. Unlike spoken language, which human children acquire effortlessly through immersion, written language must be explicitly taught. And the teaching that works — that reliably produces readers — is systematic, sequential, and dependent on a child doing hard things in a specific order.
That order begins with phonemic awareness: the ability to hear and manipulate the individual sounds within words. It proceeds through phonics: the mapping of those sounds to letters and letter combinations. It requires a child to decode — slowly, effortfully, sometimes painfully — words she has never seen before. This effortful decoding is not a problem to be solved by a more engaging interface. It is the mechanism by which the brain builds the neural pathways that eventually allow reading to become automatic. The struggle is the process. Remove the struggle and you remove the learning, no matter how colorful the animation that replaces it.
There is a term in cognitive science for the difficulty that produces durable learning: desirable difficulty. The psychologist Robert Bjork, who has spent decades studying how memory and skill are actually formed, demonstrated that conditions which feel harder in the moment — spacing out practice, varying the context, forcing retrieval without hints — produce far stronger long-term retention than conditions optimized for smooth, frictionless performance. We learn better when it is harder. We remember more when we have to work for it. This is not an opinion or a philosophy of education. It is a description of how human memory is physiologically encoded.
The educational app is, almost by definition, an engine for eliminating desirable difficulty. Its entire commercial logic depends on the child not wanting to stop — which means the child must not be frustrated, must not be confused for too long, must experience a near-continuous sense of competence and reward. The app is optimized for engagement. Engagement and learning are not only different things; in this context, they are frequently in opposition. The smoother the experience, the less the brain has to work. The less the brain works, the less it retains. The interface that feels most educational is often the one doing the least teaching.
None of this is secret knowledge. The research is not buried in obscure journals. The science of reading has been covered in major newspapers, debated in state legislatures, and incorporated — slowly, unevenly, over significant institutional resistance — into the curriculum decisions of school districts across the country. The concept of desirable difficulty appears in popular books on learning and memory that have sold in the millions. The evidence is available to anyone who looks for it.
The question is why it so rarely reaches the room where the purchase order for the educational iPad app gets signed.
Part of the answer is that the research describes a process that is slow, resistant to easy measurement, and unglamorous in its demands. It tells us that children need to struggle, that parents need to tolerate watching them struggle, that teachers need to be trained in approaches that are less intuitive than they look, and that the results will not be visible on a dashboard for months or years. This is a difficult thing to sell. It does not demo well. It does not generate the kind of enthusiasm that gets shared on a school's social media account.
What demos beautifully is a child tapping at a glowing screen with evident enthusiasm, earning stars, progressing through levels, doing — as her father is told, as she herself believes — math.
The app has not taught her mathematics. It has taught her something more immediately useful to everyone in the transaction except her: it has taught her to look like she is learning.
The Longer Curve
Bennett will grow up. The kindergarten iPad cannot see that far. It cannot see the longer curve — the one that begins with a six-year-old tapping at coins on a screen and ends, if the logic of substitution is left unchecked, with an adult who has lost the stamina to think a hard thought to its conclusion without help.
But before we arrive at that adult, it is worth being honest about something: the kindergarten iPad did not invent this logic. It inherited it. The systematic removal of cognitive friction from public life did not begin with EdTech, and it will not end with AI. It is a pattern that has been compounding for decades, across every medium through which we consume information and form ideas, and most of us have been living inside it so long that we have stopped noticing it is a pattern at all.
Consider what happened to the news. The long-form newspaper article — the kind that required a reader to hold an argument across multiple columns, to weigh evidence, to sit with ambiguity before reaching a conclusion — did not disappear because readers stopped caring about the world. It was displaced, first by cable television, which discovered that conflict and speed were more engaging than depth and nuance, and then by talk radio, which found that outrage was more engaging still. The unit of public discourse shrank from the argument to the soundbite, from the soundbite to the chyron, from the chyron to the hot take. Each compression was sold as democratization. More people, more voices, more access. What it actually produced was a culture increasingly unable to tolerate the pace at which serious thinking moves.
Social media accelerated this further and made it personal. Where cable news had shortened what we consumed, social media shortened what we were expected to produce. The op-ed became the post. The post became the caption. The caption became the reaction — a like, a share, a five-word comment fired off in the thirty seconds before the feed refreshed and carried the moment away. Platforms engineered for maximum engagement discovered, quickly, that the content most likely to be engaged with was the content least likely to require sustained thought: the outrage, the joke, the conflict, the image that needed no explanation. Depth was penalized not by design but by arithmetic. It simply moved too slowly to compete.
Short-form video completed what social media began. Where the documentary had asked a viewer to follow a complex subject across an hour or more, building understanding incrementally, the sixty-second clip begins at peak stimulation and ends before the mind has been asked to hold anything long enough to examine it. The format is not neutral. It trains the attention to expect immediate reward and to abandon anything that does not provide it. Watched habitually, across years of development, it does not merely reflect a shortened attention span. It produces one.
This is the pattern: each new medium shortened the unit of attention the culture expected of itself, and each was adopted widely enough that the shortened unit became the new baseline. A person who grew up reading long-form journalism finds cable news shallow. A person who grew up on cable news finds a newspaper article slow. A person who grew up on social media finds cable news exhausting. The baseline does not hold. It drifts, in one direction, and each generation inherits a slightly impoverished version of the cognitive expectation that preceded it.
The kindergarten iPad fits into this pattern precisely. It is not an anomaly. It is the application of thirty years of accumulated cultural logic to the earliest possible stage of human development. If friction is the enemy in the news feed, if friction is the enemy in the social media post, if friction is the enemy in the sixty-second video — why would friction be welcome in the kindergarten classroom? The EdTech industry did not have to argue hard for the removal of cognitive difficulty from early education. The culture had already decided that cognitive difficulty was a problem to be engineered away. The iPad was simply the delivery mechanism for that decision, applied to six-year-olds.
And now, at the far end of this curve, stands artificial intelligence. Every prior technology in this sequence shortened what we consumed. AI does something categorically different: it shortens what we produce. It does not merely pre-digest the world for us to receive. It offers to generate our responses to that world on our behalf. Ask it to draft the email, and we do not write the email. Ask it to structure the argument, and we do not build the argument. Ask it to summarize the document, and we do not read the document. The substitution is now total: not just consumption but expression, not just what enters the mind but what emerges from it.
Writing is not a delivery mechanism for thought. Writing is how thought gets made. When we sit down to compose a difficult email — to a superior, to an estranged friend, to someone who deserves a hard truth — and struggle to find the right words, we are not struggling because we lack efficiency. We are struggling because the thinking is genuinely hard, because the relationship between what we mean and what we can say is not yet resolved, and because the act of resolving it on the page is the act of resolving it in our minds. The AI that produces a polished draft in four seconds has not helped us think. It has thought for us, and handed us the output, and now we send words into the world that did not pass through the crucible of our own cognition.
What is being lost — across the full arc from the cable news chyron to the kindergarten math app to the AI-drafted email — is something that does not show up on any engagement metric or productivity dashboard: the accumulated capacity that comes from doing hard cognitive work repeatedly, over time, without relief. The brain is plastic. It changes in response to what it is asked to do. A mind regularly asked to read difficult prose, hold an argument across multiple pages, construct a position from incomplete evidence — that mind is being built by those activities, in a literal neurological sense. A mind that is not asked to do them is not being built. It is being serviced. And there is a significant difference between a mind that has been built and a mind that has merely been kept running.
We are now moving very fast toward a world in which almost none of this building will be necessary. The AI will write, summarize, analyze, and explain. The platform will serve us content precisely calibrated to require nothing of us. The app will make the math feel effortless. The human will tap, scroll, prompt, and approve. This is being sold to us — at every stage, for every age — as liberation. Freedom from the tedious. Elevation to higher-order thinking. More time for what matters.
It is worth asking, with some urgency, what higher-order thinking means in a person who has stopped practicing the lower orders. The capacity to synthesize depends on the capacity to read. The capacity to evaluate an argument depends on the capacity to have built one. The capacity to think clearly about complex things depends on years of having been asked to think clearly about complex things, without assistance, in the friction-filled way that actually builds the neural architecture that makes clear thinking possible. You cannot skip to the higher floor. The elevator does not exist. There are only the stairs, and we have been telling ourselves and our children, for a generation, that taking them is a waste of time.
We do not notice the atrophy because the tools keep working. The email gets sent. The report gets summarized. Somewhere, a six-year-old's coins get collected and stars accumulate.
It is only when the tool is taken away — or when the task exceeds what the tool can do, or when the moment requires a quality of thought that cannot be outsourced — that the absence of the thing that was never built becomes visible.
By then, it is very late to start building it.
The Choice
What is at stake is not, in the dramatic sense, civilization. There will be no moment of collapse, no single event that historians will later mark as the day the lights went out. What is at stake is something quieter and in some ways more consequential: the cognitive infrastructure on which everything we value about civilization depends. Democracies do not require merely that citizens show up and vote. They require citizens capable of evaluating competing claims, tolerating the ambiguity of complex problems, reading an argument to its conclusion before forming a reaction, and holding their own convictions up to scrutiny. Science does not require merely that researchers run experiments. It requires minds trained to sit with uncertainty, to follow evidence into uncomfortable places, to resist the pull of the answer that arrives too easily. Culture does not require merely that people consume art. It requires people with the patience and inner resources to be changed by it.
None of these capacities are guaranteed. None of them are natural. Every one of them is produced — slowly, effortfully, across years of education and practice — by exactly the kind of cognitive work that we have spent the last three decades engineering out of daily life. The forms of civilization can persist long after the substance has hollowed out. The elections continue. The journals publish. The classrooms fill. But if the habits of mind that animate those forms are quietly degrading in the population that operates them, the forms become theater. Convincing theater, perhaps. Theater that can run for quite a long time before anyone in the audience thinks to ask whether anything real is happening on the stage.
In 2006, the filmmaker Mike Judge released a comedy called Idiocracy. The premise was blunt to the point of rudeness: a perfectly average man wakes up five hundred years in the future to discover that humanity, having selected for generations against intelligence and in favor of whatever was easiest and most immediately gratifying, has become comprehensively, catastrophically stupid. The film was not a hit. It was dismissed by many critics as mean-spirited and overly broad. It has since become one of the most-quoted films in the English language, cited so frequently as a description of current events that the references have themselves become a cliché.
What Judge understood, beneath the broad comedy, was the mechanism. Idiocracy is not really about stupidity. It is about drift. It is about what happens when a civilization makes, at every level and in every generation, the small and individually reasonable decision to prefer the easier path. No single decision is ruinous. No single generation fails catastrophically. The drift is gradual, compounding, and very difficult to perceive from inside it because each generation inherits the diminished baseline as normal and has no felt sense of what was lost. The film played it for laughs because the alternative — playing it straight — would have been too uncomfortable to watch. It is considerably more uncomfortable now.
Judge set his film five centuries in the future. He did not need to. The mechanism he identified is not a distant risk. It is a present condition, visible in the things we have already discussed: in the public discourse that can no longer sustain a complex argument, in the classrooms that have replaced the struggle of learning with the sensation of it, in the adults who have quietly outsourced the hard work of thinking to tools that are happy to oblige. The timeline is not five hundred years. It is the thirty years already behind us and the thirty years immediately ahead.
And yet.
There is a reason this essay exists, and it is not despair. Every generation that has faced a medium or a technology that threatened to degrade the quality of public thought has also produced people who understood the threat and responded to it — not by rejecting the technology wholesale, but by developing a conscious relationship with it. The printing press alarmed scholars who feared it would flood the world with bad ideas. It did. It also produced the pamphlets of the Enlightenment. Television was going to rot every mind it reached. It did damage. It also produced journalism, documentary, and drama of enduring seriousness. The tools are never only one thing. What determines what they do to us is whether we are paying attention.
The unconsciousness is the enemy. Not the iPad, not the algorithm, not the AI — but the failure to look clearly at what these things are doing to us and to our children, and to make deliberate choices in response. The EdTech industry thrives on incuriosity. The attention economy depends on passivity. The AI assistant requires only that we keep prompting. None of these systems need us to be stupid. They need us to be unquestioning. And that is a different problem, because it is one we can actually do something about.
Which brings me back to Bennett.
He is six years old. He is learning to read and to count and to understand the world, and the habits of mind he builds in the next few years will shape the architecture of his thinking for the rest of his life. I cannot fix the EdTech procurement process. I cannot redesign the attention economy. I cannot make the AI less capable or the platforms less engineered for capture. What I can do — what any parent can do, what any teacher willing to push back against institutional inertia can do — is refuse to accept the substitution without examination. Ask what the app is actually teaching. Demand outcomes rather than engagement metrics. Let him be frustrated by a problem long enough for the frustration to become understanding. Put a pencil in his hand.
None of this is heroic. It does not require a manifesto or a movement or a school board confrontation, though those things have their place. It requires only the willingness to notice what is happening and to take it seriously — to resist the pull of the frictionless, the optimized, the endlessly engaging, on behalf of something slower and harder and more durable.
The atrophy is not inevitable. It is a direction, not a destiny. But it moves quietly, and it moves fast, and it has a very significant head start.
The first thing — the only necessary first thing — is to notice it.
That is what Bennett's iPad taught me, on an otherwise unremarkable Tuesday evening, when my son slid a screen across the kitchen table and told me, with complete and untroubled confidence, that what he was doing was math.
He was wrong. But he was six, and he had been told otherwise by people he trusted, and the game was very good at feeling like the real thing.
That is precisely what makes it dangerous.
And precisely why we have to pay attention.