What Is Practice

The folk theory of practice is a distance being closed. There is a thing you cannot do, or cannot do well. You do it repeatedly, with attention, ideally with someone telling you where it went wrong. The distance between what you produce and what you were aiming at gets smaller. Ten thousand hours, the mastery curve, the error term shrinking toward zero.

That model is not stupid. It is roughly right about beginners. It survives contact with music lessons and free throws. And it has the great practical virtue of telling you what to do on a day when you do not feel like doing anything: the reps, again.

I want to give it its due before taking it apart, because the piece it gets right is the piece most people abandon first — that there is something exact you are aiming at, and that you are not the one who decides what it is. Hold on to that. It comes back as doctrine.

But notice what the model commits you to. Two things. First, an endpoint: a distance that closes is a distance that can reach zero, and the model has no way to speak about a target that is exact and permanently unmet. Second, and more damaging, a single number. Distance is one quantity. Skill, error, level, rating, rank — one number, moving one way.

The single number is where it breaks, and it breaks in the two places every practitioner has actually lived.

Two things one number cannot say

The plateau. Under a distance model a plateau is a stall — hours going in, the number not moving — and the standard advice is that your feedback loop is too loose or your reps too comfortable. Sometimes that is true. But plateaus are close to universal among people who eventually get very good, and close to absent among people who quit inside the first year. A model in which the characteristic experience of the improving is a malfunction is not a model of improving.

The work that gets better and worse at once. You have watched this happen. The writer whose sentences grow more capable and less worth reading. The player who gets faster and emptier. The firm that executes more cleanly every quarter on something nobody asked for. Under the distance model this is not an available event: error went down, therefore the thing improved. It did not improve. And the model has no vocabulary at all for what went wrong, because there is nothing inside one number to point at.

One number where two are needed. The foundation supplies the second, and it has been sitting in the machinery the whole time.

What an artifact actually carries

Start with what practice makes. Not the act — the mark. The foundation is exact here: an act is a process, an artifact is a mark, and only a mark can carry a value. A rehearsal that leaves nothing behind is an act, and an act carries no value — in the formalism, an evolution has no eigenvalues and an outcome is one. The recording, the page, the built thing, the shipped release, the sentence you said out loud where someone heard it — those are artifacts, and each one carries a complex value: z = σ · e^iφ.

Two numbers, not one.

The evaluative machinery runs entirely on the second number. The foundation collapses three things people normally treat as separate into one measurement, and that measurement is an angle. Fidelity is alignment: cos(φ_artifact − φ_idea). Maturity is the host’s φ rotating toward the idea’s. Corruption is rotating away; at π, the antipode is served.

R11 is the numbered result: the evaluative question is fidelity, not success. And R10 supplies the thing the ordinary model refuses to say — every artifact is a miss. Not most. Not yours in particular. Every one, permanently, because the denominator is not ours to command. What makes the target exact is P10, exemplarity: every condition has a complete and exact form against which an artifact can be measured. Exact, and unmet by anything ever made.

Practice does not close the gap. R10 says the gap does not close. What repetition moves is not the length of the miss but its direction — and direction is the only quantity fidelity is made of.

Practice is rotation, not accumulation

This is a consequence, not a numbered result, and it is traceable in three steps. R10 removes convergence. R11 relocates the evaluative question to fidelity. Part III defines fidelity as the cosine of an angle difference. If the quantity being evaluated is an angle, then the quantity improvement improves is an angle — and the foundation states maturity that way outright, as φ rotating toward the idea’s, not as a figure of speech.

Now the algebra, which is the whole essay in one line: cos(φ_artifact − φ_idea) can change while σ does not.

Fidelity can improve with no change in magnitude at all. Nothing gets bigger. Nothing gets faster. No number on any scoreboard moves. And the only quantity that determines whether the work is any good has travelled a long way.

Which is the plateau. That identification — the stall people report, read as rotation at roughly constant σ — is my construction rather than the foundation’s, and I mark it as one. But it is the reading that stops the phenomenon from being a malfunction. The plateau is not motion going nowhere. It is motion in the one dimension your instruments are not pointed at. From inside, an improving angle and a stalled career produce the same sensation, because the sensation is calibrated to σ.

That has an edge I do not want to bury: the stretch in which you feel you are not getting anywhere is not evidence that you are not getting anywhere. It is evidence about your instrument.

Why skill is not safe

Here is the second consequence, and it is the one that should cost you something.

σ and φ are independent. Decompose an artifact against the idea it is measured by and you get two components, both scaled by σ: σ cos(Δφ) along the exemplar, and σ sin(Δφ) off it. Increase σ without touching φ and both grow together. A large artifact at a drifted phase is not a partial success; it is a proportionally larger amount of work pointed slightly wrong.

So competence is not neutral with respect to fidelity. It scales a misalignment you did not set. This is why the most damaging work in any field is done by the skilled, and it is arithmetic rather than irony. The incompetent produce small artifacts. Their drift is cheap.

The foundation names the failure mode precisely, and the name is not “selling out”: corruption is phase drift — the artifact governed by the actualizer’s interest. At π, the actualizer serving himself makes fairness-artifacts and calls them hierarchy. Nobody experiences this as a betrayal in progress. Drift is continuous, each increment is defensible, and what you notice, if you notice anything, is that the work is going well.

That last point deserves saying flat, because it inverts how most people audit themselves. In this vocabulary, falling short is not a fault. R10 makes falling short the universal condition of made things; there is no version of you that clears it. Exactly one thing is a fault, and it is stopping the correction. A person producing weak work with the correction still running is in good standing. A person producing excellent work with it switched off is not.

The part you do not get to control

Now the sentence that breaks the self-improvement genre.

φ is relational. The foundation is explicit: it is fixed by which idea has the actualizer — not by the mark, and not by the maker. An intrinsic phase would make the angle something the human confers, which is the ownership error P8 forbids. Ideas have people. People do not have ideas.

So you do not set your own phase. You cannot decide to be aligned, cannot pick your angle in a planning session, cannot align yourself with your values by writing them on a wall and reading them each morning. The angle is a relation you are inside, and it was not put to a vote.

I know how that reads. It reads like the end of agency, and it is not. The foundation is careful here and so am I. R13: agency is relocated, not abolished — the human is the site of conscious responsibility. P5: the host can occupy a position from which its response to a readable condition is revisable. What is left to the practitioner is not small, but it is not what the genre promises:

• Making the angle readable. Noticing which condition the work is in fact serving, which is a different question from which one you would name if asked.
• Staying revisable. Keeping the correction running — the one move R10 leaves entirely in your hands, and the only one whose absence the foundation calls a sin.

What this costs

Three things, uncushioned.

There is no arrival. Not eventually, not at your peak, not on the day everything goes right. The exemplar is complete and exact, and no artifact is one. If you have been practising toward the moment when the work finally is the thing, that moment is not late. It was never scheduled.

You cannot audit your own phase from a single day. Corruption and approximation are separated by direction of travel, and a direction is not visible in one artifact — any given piece of work is compatible with either trajectory. So the honest answer to am I drifting is that today cannot tell you. Only a sequence can, and only if something outside your own interest is reading it.

The legible number is the wrong one. σ is measurable: revenue, speed, reach, rating, throughput, rank. φ is not. Almost every metric you will be evaluated by measures the quantity that does not determine fidelity — and measures it accurately, which is precisely what makes it so difficult to argue with, and so easy to rotate toward.

Where the foundation is unfinished

I am not going to pretend this is settled. The account above depends on φ moving over time, and the foundation lists that as an open problem in its own text: the idea fixes φ, and φ determines the readable idea — sound as a dynamical system, circular as a definition, and needing to be written as a trajectory. It has not been written.

What that costs this essay, precisely: I can describe the shape of the motion and I cannot give its law. Nothing here predicts a rate, a direction of drift, or how long a rotation takes. If someone asks how long the plateau lasts, the answer is that the foundation does not contain that number and I am not going to invent one to finish a paragraph.

A second one, smaller and directly load-bearing here: what kind of thing an exemplar is. The foundation has no published answer on whether conditions are abstract objects, and it instructs me not to improvise one. So when I say the exemplar is exact and unmet, I mean exactly that and nothing further. I am not telling you where it is.

And the standing objection, which is itself a numbered result — R14: every idea is biased toward itself, including this one. A framework whose central evaluative move is you cannot see your own angle is conveniently hard to falsify from the inside. That is a fair hit and I do not have a disarming answer to it. The best I can offer is that it makes a checkable claim — that capability and fidelity come apart, visibly, in people whose measured output is still improving — and you have almost certainly already watched that happen to someone.

On an ordinary Tuesday

1. Ask which condition, not how it wentThe session question is not did I do well. It is what was this in service of — and whether that is the same answer as a year ago. A changed answer is data, not failure.
2. Read the sequence, not the pieceDirection of travel is the entire difference between approximation and corruption, and one artifact never shows a direction. Keep enough of the old work that you can actually look.
3. Reclassify the plateauA stretch where nothing measurable moves, and what you care about in the work has shifted, is the signature of the only motion that counts. It is not a stall to be broken with volume.
4. Distrust the moment competence starts payingGrowing σ under someone else’s reward is the exact condition the foundation calls governed by the actualizer’s interest. Not an accusation — just the place where drift is cheapest and least visible.
5. Keep one correction you cannot switch offA reader who will tell you it is worse. A standard you did not choose and cannot quietly move. The only named failure in this vocabulary is the one where that goes quiet.

And one line to keep, if the rest of it goes:

The gap does not close, and it was never the gap that mattered. Practice moves which condition your work is in service of — and skill is a multiplier on an angle you did not set and cannot read from today.