Downloading Kung FuWhy You Can't Read Your Way to Understanding

You are about to read an essay arguing that reading is a terrible way to understand things. I’m aware of the irony. You’re sitting with a piece of text that is trying to convince you, through text, that text doesn’t work.

I can’t resolve this contradiction for you. What I can do is point at something and hope you go look at it yourself. Think of this essay as a flare, not the fire. To actually understand what I’m talking about, you will eventually need to stop reading and go do something difficult.

Which brings us to The Matrix.

You remember the scene. Neo reclines in a chair. A cable jacked into the base of his skull feeds data directly into his brain. He opens his eyes. “I know Kung Fu.”

Notice your own reaction to that scene. When you imagine what just happened to Neo, you do not imagine that he read about kung fu really fast. You do not picture a PDF on the history of martial arts scrolling past at superhuman speed, or a seminar on biomechanics playing at 10,000x. You picture something categorically different from reading. You picture him knowing. Having it in his body, in his reflexes, in the way he moves.

That instinct is worth paying attention to. The feeling of knowing something and the feeling of having read about something are not the same feeling. They are not even close. When Neo says “I know Kung Fu,” nobody in the audience thinks “oh, he read about kung fu.” They think “he has kung fu.” Those are different things. And you already knew that before I pointed it out.

But notice something else. Neo can fight. He has the reflexes, the technique, the speed. What the movie never shows is whether he can think about martial arts. Can he watch two other fighters and see what they are doing wrong? Can he predict what an opponent will do three moves from now? Can he be surprised by something unexpected and learn from the surprise? The download gave him performance. What it skipped is the thing that matters most: the capacity to reason about the domain.

So here is the question. If you already know, intuitively, that reading about a thing and knowing a thing are fundamentally different, why do we keep building our entire educational and professional culture around reading and discussion, as though reading about a thing could teach you to reason about it?

The Poverty of Text and Talk

There is a distinction in philosophy that gets at this directly. In The Concept of Mind, the philosopher Gilbert Ryle called it the difference between “knowing that” and “knowing how.”¹

But notice what this distinction does not mean. It does not mean you should trade knowing-that for knowing-how and stay there. The relationship runs the other direction. Knowing how to do a thing is what eventually lets you think about it at a level that reading alone never reaches. You don’t practise to become a permanent practitioner. You practise to reach the point where real thinking becomes possible. And not just any practice. Practice where you are corrected. Where you do the thing wrong, find out immediately, and adjust. That correction is the mechanism. Without it, repetition is just repetition.

Say you want to learn to write well. You can read books on rhetoric and argument structure. You can attend workshops where published authors explain how they build an essay. You will come away with vocabulary, frameworks, and a genuine appreciation for how good writing works. But the first time you sit down to write something of your own, the blank page does not care what you read about structure. Your sentences come out tangled. Your argument drifts. The gap between what you know about good writing and what you can produce is enormous.

But grind through a few dozen drafts. Write and rewrite and throw out and start over. Each draft tells you what you got wrong. The argument that felt airtight in your head collapses on the page. You rewrite it. It collapses differently. You rewrite it again. Something shifts. You start seeing the shape of an argument before you have written it down. You can read someone else’s essay and predict where it will fall apart, even before you can articulate why. You are not consulting your notes on rhetoric. You are thinking about writing. Not performing opinions about it. Actually reasoning about structure, argument, and audience in real time. That capacity did not come from reading about writing. It came from the writing, and from every draft that showed you where your reasoning broke down.

The same pattern holds with mathematics. You can read a textbook on integration. You can attend a lecture series. You can have a lively group discussion about the intuition behind the fundamental theorem. But the first time you sit down alone with an integral and a blank page, none of that helps you. The notation swims. You can’t tell which technique to apply. The problem doesn’t care how articulate you were in the discussion.

But work through a hundred problems and something else happens. Not just speed. You start seeing mathematical structure. You can look at a proof and see why it works. You spot connections between theorems that used to look unrelated. You can reason about a computation without having to grind through every step. And the reason it worked is that every wrong answer was a correction. The problem told you, immediately, that your approach was off. That feedback, repeated hundreds of times, is what built the capacity to think mathematically. The mechanical work was the prerequisite to the reasoning, not the end goal.

This is the same gap you felt watching Neo. Reading about a thing and being able to think about it deeply are not two points on a continuum. They are different categories of knowledge. And the only path from one to the other runs through corrected practice: doing, failing, being told what you got wrong, and adjusting. This is not specific to writing or mathematics or chess. It is how expertise works in every domain. Every domain has boundaries between what counts as participating in the subject and what falls outside it. Correction is what shows you when you have crossed that line. The boundaries differ from field to field, but the mechanism does not. And the higher-order thinking that corrected practice builds is about the domain you practised in, not a general-purpose upgrade to your brain. Being a chess grandmaster does not make you a better doctor. Corrected practice is what builds the capacity for higher-order thinking. The ability to reason about a domain. To predict what will happen. To be surprised when something breaks the pattern.

(I know this reads like I’m heading for “just practise more.” I’m not. Mindless repetition doesn’t build anything. What builds the capacity for real thinking is practice at the edge of your ability, where you are being corrected. The interesting part is what that process does to your brain.)

A chess grandmaster does not play brilliant chess because she has read more books on chess theory than you. She plays brilliant chess because her brain has been physically restructured by thousands of hours of playing, losing, and adjusting. Hand her and a novice the same position and she does not just think about it differently. She sees a different board. The novice sees thirty-two pieces. She sees three interlocking threat structures and a tempo advantage she can convert in six moves. She can predict her opponent’s likely responses three moves deep. She is surprised when something breaks her expectation, and that surprise tells her something important about the position. This is not memorization. This is higher-order reasoning about chess, and it was built entirely by corrected practice. Adriaan de Groot confirmed this in the 1940s: show a master a chess position for five seconds and she reproduces it almost perfectly. Show her a random arrangement instead and she does no better than a beginner.² The expertise is not memory. It is the patterns.

You cannot read your way into that. You cannot discuss your way into it. It is the residue of doing and failing and being corrected, and no amount of reading closes the gap.

The Architecture of Real Understanding

So what is practice actually doing? This is where the case stops being philosophical and starts being biological.

Say you’re learning to program for the first time. You’re trying to write a function, and you have to simultaneously hold in your head: what a variable is, what a function does, what the syntax for a loop looks like, what you’re trying to compute, and how the pieces fit together. It’s overwhelming. Your brain is full before you’ve even started on the interesting part.

This is not a failure of intelligence. It’s a constraint of architecture. Human working memory, the mental workspace where you hold and manipulate information in real time, is limited to roughly four to seven items.³ That’s it. Four to seven things, held in a buffer that starts decaying within seconds if you stop attending to them.

A programmer who has written a thousand functions does not think about what a variable is. She does not think about loop syntax. Those things have become automatic. Her brain has compressed the individual components into single, retrievable patterns. George Miller called this “chunking.” Where the novice sees a dozen separate pieces, the expert sees one familiar shape.

This is the mechanism behind everything I’ve been describing. Chunking is how your brain clears the workspace for higher-order thinking. Until the basics are automatic, all your cognitive resources are consumed by the mechanics. There is nothing left over for the interesting work: reasoning about the problem, predicting what will happen, noticing when something is off.

To be clear: reading is not useless. Reading is how you learn the steps of the algorithm, the rules of the notation, the structure of the proof. Among people who already share the higher-level encodings, text is how ideas travel: two experts can exchange insights through writing precisely because they have the chunked representations to decode each other. And for practitioners with enough experience, reading itself becomes a form of corrected practice: they evaluate what they read against their own patterns and notice when something does not fit. But what reading cannot do is build those encodings in the first place. And until they are automatic, you cannot reason with them. You can only laboriously execute them one at a time while your working memory gasps for air. It is like a pianist who is still searching for the right keys. She cannot hear the music because all of her attention is on her fingers. The romantic notion that you can skip the mechanical grind and jump straight to creative thinking misunderstands what creativity requires. It requires that the mechanics be invisible.

This is the part that people get backwards. Drills, repetition, and correction are not the enemies of creativity. They are the prerequisites. The higher-order reasoning everyone wants to skip to is only accessible after the lower-level components have been chunked. You are not practising to become a robot. You are practising so that your brain has the cognitive space to think.

And the destination is not the mechanical skill itself. A composer may not perform at concert level, but she could not compose without having struggled through the instrument. The best mathematical reasoners are not necessarily great at arithmetic, but they ground through it. The correlation between the mechanical skill and the higher-order thinking it unlocks is low. The dependency is absolute.

This is not just an argument from first principles. The psychologist John Hayes analysed 500 notable musical compositions by 76 composers and found that virtually none were produced before the composer’s tenth year of serious work.⁴ Even Mozart. Hayes called it the “ten years of silence”: a decade of building the chunks before the creative work that mattered became possible.

But chunking does not happen through repetition alone. It happens through corrected repetition. You write the function and it breaks. You fix it and it breaks differently. You fix that and it works but it’s slow. Each cycle, your brain encodes a little more. Each correction is the signal that drives the encoding. Without the feedback, you are just repeating your mistakes.

When this process runs long enough, something changes that goes beyond simple competence. The novice who has only read about a subject sees a flat, theoretical landscape. Everything looks roughly the same. After hundreds of hours of corrected practice, the practitioner begins to see patterns that were always there but were invisible before. But it goes further than seeing. An experienced clinician does not just perceive something a medical student cannot see. She can reason about it. She can predict how a condition will progress. She notices when something does not fit the expected pattern, and that surprise tells her something diagnostically important. A seasoned engineer does not just feel where the failure point is. He can reason about why it will fail and what will happen when it does.

This is not a metaphor. The expert’s brain has been physically restructured by corrected practice. The patterns are encoded in neural pathways that only form through the specific, grinding process of doing and failing and adjusting. Reading cannot build those pathways. Discussion cannot build them. They are what corrected practice leaves behind, and they are the biological foundation of higher-order thinking. Neuroscience has measured this directly: years of corrected practice produce measurable changes in brain structure. London taxi drivers who spend years memorising 25,000 streets show measurable hippocampal growth; those who attempt the same training and fail show none.⁵ The practice literally builds the structure that makes the reasoning possible.

The Evidence

I’ve been making claims about how expertise works. Let me back them up with something more durable than argumentation.

The psychologist Anders Ericsson spent his career studying expert performance. Not by surveying experts about what they thought made them good, but by putting experts and novices side by side in controlled settings and measuring the differences.

What Ericsson found is that the kind of thinking I’ve been describing, the higher-order reasoning that separates experts from amateurs, is built through what he called “deliberate practice.” The definition is specific and worth unpacking, because each part of it is a test that reading and discussion fail. Deliberate practice must be targeted at identified weaknesses, not vague exploration. It must be at the edge of your current ability, in the zone where failure is likely. And it must involve immediate, objective feedback that tells you exactly what you did wrong.

The evidence is consistent. In study after study, the gap between the best and the good comes down to hours of solitary, corrected practice: not hours in seminars, not hours discussing technique, but hours alone with the problem, getting things wrong and adjusting.⁶

Test a seminar against those criteria. A participant in a discussion about, say, financial analysis cannot “deliberately practice” financial analysis by talking about its principles. The practice is not targeted at weaknesses. It’s general and abstract. It is not at the edge of ability. There is no measurable challenge. And there is no real correction. Just the diffuse social signals of head-nods and polite disagreement. By Ericsson’s criteria, a seminar doesn’t just fall short of deliberate practice. It fails to qualify as practice at all.

But the clearest illustration of what higher-order thinking looks like, and why reading cannot build it, comes from a different line of research.

Imagine you’re watching a video of a tennis serve. The server tosses the ball, begins the motion, and right before the racket makes contact the video pauses. Black screen. What happens next? Where is the ball going?

If you’re a novice, you guess. You perform at or near chance. You have no basis for prediction.

If you’re an expert tennis player, you know where the serve is going before the racket makes contact. You can read it from subtle cues in the server’s shoulder rotation, toss trajectory, and weight distribution. Cues that the novice literally cannot see, because seeing them requires thousands of hours of having played points and experienced their outcomes. Not thousands of hours watching. Playing. Getting the return wrong, seeing where the ball actually went, and adjusting.

Researchers call this the temporal occlusion paradigm. It has been replicated across dozens of sports since the late 1970s,⁷ and the results are not a difference of degree. They are a chasm.

This is higher-order thinking in action. The expert is not just reacting faster. She is reasoning about the future of an action in real time, drawing on patterns that corrected practice burned into her neural pathways. She can predict what will happen next. She is surprised when a serve breaks the expected pattern, and that surprise tells her something important. This is the capacity that chunking builds. This is the payoff of all those hours of doing, failing, and adjusting.

A novice could read a hundred books on tennis biomechanics. She could lead a dozen seminars on serve-return strategy. She could produce a lovely slide deck on anticipatory perception in racket sports. She would still fail the test. She would still be unable to think what the expert thinks. Because the expert’s advantage is not informational. It is cognitive architecture built by corrected practice, and it lives in neural pathways that reading and discussion cannot form.

You cannot read your way into this kind of thinking. You cannot discuss your way into it.

The Ego Shield and the Venal Rewards

So if reading and discussion are such poor substitutes for the real thing, why have we built our entire educational and corporate infrastructure around them?

Part of it is practical. Discussion is scalable. You can put forty people in a room and have them talk about leadership for an hour. It’s cheap. It doesn’t require individual assessment. But the practical explanation only goes so far. There is something deeper going on, and it has to do with what discussion actually rewards.

I have been in hundreds of meetings and seminars where people talk about a subject instead of doing it. I have been the person who uses the right jargon at the right moment, who projects confidence, who steers the conversation in a way that makes me look sharp. It feels great. You learn nothing. What you walk away with is the pleasant feeling that your opinion mattered. That you contributed something. I’d call these venal rewards, because they have nothing to do with whether anyone in the room actually got better at the subject under discussion. They are rewards for performing competence, not for developing it.

Not all engagement with a subject is performance. Writing is corrected practice: the draft tells you, concretely, what you got wrong. Your argument collapses on the page and you have to rebuild it. Formal debate is corrected practice: your position gets dismantled and you find out exactly where your reasoning broke. A casual seminar discussion is neither. A student raises their hand and offers a clever interpretation. The professor nods. Other students look impressed. The student walks out feeling like they understood the material. But they never tested that understanding against anything. What they practised was not the subject. It was the performance of having opinions about the subject. Writing and debate reward the willingness to be shown where your thinking failed.

Compare this to what happens when you engage directly with the subject itself. It doesn’t have to be solitary. A teacher correcting your form, a chess opponent punishing your mistakes, a debater dismantling your argument, a draft that collapses under its own weight: these are all corrected practice. What matters is not whether you are alone but whether the feedback tells you what you got wrong. There is just you and the thing, and the thing does not care how articulate you are. When it tells you that you are wrong, the correction is immediate, concrete, and impossible to spin.

This is the part nobody wants to talk about. Being wrong is painful. Not painful in a vague, philosophical sense. Painful the way embarrassment is painful. You feel it in your chest.

And here is what makes it worse: if you’re practising at the edge of your abilities, which is the only way the mechanism works, you are getting things wrong constantly. Not occasionally. Not as a rare humbling event. Constantly. Every session. The whole point of practising at the edge is that you are always slightly beyond what you can do, which means you are always failing, which means you are always being corrected. The correction is not a setback. It is the mechanism. But it doesn’t feel like a mechanism. It feels like being told, over and over, that you’re not good enough.

There is a second emotional barrier that gets less attention. Following a prescribed process, working through directed steps, submitting to an algorithm you didn’t choose: this triggers something that feels like being bossed around. Children feel it acutely. Adults feel it too but dress it up differently. “I don’t learn that way.” “I need to understand the why before I follow the steps.” “Rote learning kills creativity.” Sometimes these are genuine learning preferences. Often they are the emotional aversion to surrendering control, dressed up as pedagogy.

The irony is that the surrender is temporary. Once the fundamentals are automatic, you have more freedom than you started with. The pianist who submitted to years of scales can now improvise. The programmer who ground through hundreds of exercises can now architect systems. The mechanical submission is the price of cognitive freedom. But in the moment, it does not feel like freedom. It feels like obedience.

These two emotional barriers, the pain of correction and the aversion to following prescribed steps, do not just make people avoid practice. They make people opponents of it. Half the people you know will tell you, unprompted, that they are “not math people.” This is treated as a fixed fact about their brains, as innate as eye colour. In most cases, it is not. It is the product of a system that failed to teach properly: wrong lesson, wrong level, wrong time. The student fell behind, the curriculum moved on, and they concluded they were fundamentally incapable. Once you believe that, correction stops being feedback and starts being punishment. Of course you would hate testing. Of course you would take aim at it. Of course you would become an opponent of the whole exercise. The aversion is completely rational. The premise is false.

And here is where the argument has to be honest about what the answer is not. The answer is not “toughen up.” The answer is not “submit to the process and stop complaining.” If someone has been burned by correction in an environment that made them feel stupid, telling them to endure more correction is not a solution. It is cruelty.

The answer is to fix the environment. Lower the stakes. Give feedback that is immediate and specific but not punitive. Put people in front of teachers who are genuinely on their side, who treat errors as information rather than failure. Help with the emotional part: put the feelings in context. The discomfort of being wrong is a sign that you are learning, not a sign that you cannot. None of this is soft. It is the engineering required to deliver corrected practice without breaking the person receiving it.

But what institutions actually do is the opposite. They eliminate the practice altogether. Open-ended conversation is an ego shield. In a seminar, no one is ever told they are wrong. Group projects distribute accountability so thinly that individual failure becomes invisible. This is not a bug in how we’ve built schools and companies. It is the architecture, designed at every level to avoid the discomfort of being corrected. We have built our educational and corporate cultures not to optimise learning, but to protect our egos from the sting of failure. And in doing so, we have eliminated the very mechanism that builds higher-order thinking.

But there is an institutional form of corrected practice. Give a student a problem. Let her work it. Tell her what she got wrong. Give her another. This is testing: not a final exam that sorts winners from losers, but regular, low-stakes problems that tell you what you got wrong while there is still time to fix it. Most arguments against testing are arguments against testing done badly: high stakes, no second chances, measuring whether you can calculate under pressure rather than whether you can reason. The mechanical work is still the prerequisite. But the right test measures whether the reasoning arrived, not whether the student can execute the steps that built it.

The Flare, Not the Fire

Some of the best mathematical thinkers in the world want to skip the arithmetic. They remember the moment theory clicked, not the years of problem sets that made the click possible. The mechanical work disappears from memory the moment it succeeds. This is the curse of knowledge applied to learning itself: the practice works, but you will not notice it working.

This essay is no exception. It is a flare, not the fire. If anything here resonated, it is because you already had the practice to decode it.

But here is a question that blind spot makes easy to miss. How many people never got the chance? How many concluded they were “not math people” or “not analytical” because no system ever gave them the right corrected practice in the right environment at the right time?

You are not only a learner. You are also, at various moments, a teacher, a manager, a parent, a voter. That last one is easy to forget. Education policy is not something that happens to other people. You live in a democracy. The systems that failed that twelve-year-old were built by people who decided this was not their problem. Every time you design a training program, choose a curriculum, or vote for the people who set education policy, you are choosing between systems that protect egos and systems that build the capacity for higher-order thinking.

We tried to fix this once. We called it gamification. We copied the surface of games (points, badges, leaderboards) instead of the mechanism (corrected practice at the edge of ability, with immediate feedback, in a low-stakes environment). The mechanism that actually makes games effective at teaching is the one this essay has been describing. No ego shield. No venal rewards. Just you and the thing, failing and adjusting until you can think about it at a level you couldn’t before.

Picture a twelve-year-old staring at a math problem she can’t solve. She has gotten the last three wrong. She is about to decide she is not a math person. The system she is in will determine which way this goes. If it moves on to the next lesson, she closes the book. If it adjusts the problem, sharpens the feedback, and meets her where she actually is, if it makes the correction feel safe rather than punishing, then ten minutes later she solves one. Then another. She does not feel smart yet. But she has stopped believing she cannot. What the system is doing is testing her: not to rank her, but to correct her. Regular, low-stakes problems adjusted to her level, repeated until mastery. Students who are tested this way become less anxious about tests, not more.⁸ And in a year, if the system keeps meeting her where she is, she will start to reason about mathematics in ways she cannot currently imagine.

We know how learning works. We know that higher-order thinking is built by corrected practice, not by reading and discussion. We know that the emotional barriers are real and that the answer is better environments, not tougher students. The research is not ambiguous. The question is whether we will build systems that reflect what we know, or keep building ego shields and calling them education.