There’s a contradiction running through this material and it’s time to resolve it.

On one side: counting cancels the thing you’re counting. Keep a tally of the good you’ve done and you’ve converted it into a transaction, and transactions close. The man running a ledger is the same man he was, with a spreadsheet.

On the other side: during any long change, output reads zero, so you must measure inputs or you’ll quit. Measure the thing you control, write it down, or the accumulation genuinely disappears.

Both are true and they look like they can’t be.

The resolution is that there are two kinds of counting, they use identical numbers, and they differ entirely in what the number is for.

The Distinction

Recording an activity can serve either as instrumentation or as scoring, and the two are indistinguishable by the record itself. Instrumentation treats the number as a reading used to steer: it carries no claim about the person’s worth, generates no expectation of return, and its function is to inform adjustment. Scoring treats the number as either a claim about the person’s standing or a balance against which some return is anticipated. The distinction is not in the data recorded but in what the record is for, and it is testable by the response to an unfavourable reading: instrumentation produces adjustment, scoring produces shame, defensiveness or a sense of unfairness. Scoring reintroduces the transactional structure that prevents accumulation, since an action performed to increase a total has collected its return at the point of entry. Instrumentation does not, because no return is being sought.

Two kinds. One test.

The Test

Here’s how you find out which one you’re running, and it takes about four seconds.

What happens when the number is bad?

Instrumentation: you look at it, note it, and think about what to change. Mild interest. The bad reading is information, and information is what you built the instrument for.

Scoring: you feel something. Shame, defensiveness, a small internal argument about why this week doesn’t count, or a flicker of unfairness that you’ve been doing all this and it hasn’t paid off.

That feeling is the tell, and it’s completely reliable. A thermometer reading a low temperature doesn’t produce guilt. If your record does, it isn’t a thermometer.

There’s a second, subtler version of the same test:

What happens when the number is good?

Instrumentation: nothing much. Adjustment isn’t needed.

Scoring: a small lift. And that lift is the payout — which means the account just closed, and the thing you did has already been spent on the feeling of having done it. That’s the mechanism in half good and full good.

Five Rules

Practical rules that keep a record on the instrumentation side. They’re all small and they all matter.

One: Record inputs, not merit

Write what you did, not how you did.

Declined two requests is an input. Was good about boundaries this week is a merit assessment, and merit assessments are claims about you.

The difference sounds pedantic and it’s the whole thing. An input is a fact about the world. A merit assessment is a fact about your standing, and standing invites scoring.

Two: No running total

Record entries. Don’t display a cumulative figure.

A total is a score. It exists to be compared — to last month, to what you expected, to what you think someone in your position should have. And comparison is scoring by another route.

Individual entries don’t do that. A list of forty things you did is evidence. A number that says forty is a result, and results have opinions attached.

Three: Never share it

The moment a record has an audience, it becomes a claim.

You’ll edit it. Not deliberately — but you’ll round up, you’ll frame the misses, you’ll leave out the fortnight where nothing happened. A record kept for someone else is a performance, and performances discharge at the point of being seen.

This is the rule people most want to break, because sharing produces accountability. The accountability is real and the cost is that you’ve converted an instrument into a claim. If you need accountability, get it separately — a named commitment to a named person on a named date — and keep the record private.

Four: Review on a schedule

Look at it on a fixed interval. Not continuously.

Continuous checking is scoring behaviour. Nobody checks an instrument they’re using to steer forty times a day; they check it when a decision is due.

If you’re looking at the record frequently, you’re not reading it. You’re seeking something from it, and what you’re seeking is reassurance, which is a payout.

Monthly is usually right. Weekly for a short intense period. Daily is scoring wearing a tracking costume.

Five: Record the misses, flatly

A record that only contains successes is a highlight reel, and highlight reels are claims.

Write down the days it didn’t happen, in the same tone, with no commentary. No explanation, no reason, no small note about how busy the week was.

This is the hardest of the five and it’s the most diagnostic. If you find yourself unable to write a flat entry for a bad week — if you need to add context — you’re scoring, and the context is a defence.

The One That Actually Worked

Worth noting that the four-hundred-year-old case at the centre of this material involved a ledger.

Yuan Liaofan kept one. His wife kept one too. He committed to three thousand acts and he counted them, over roughly a decade.

By the argument that counting cancels, that should have failed completely.

It didn’t, and the reason is the distinction above. The ledger was a training instrument. It told him whether the practice was actually happening, over a period long enough that memory would have been useless. It wasn’t an invoice submitted to the universe, and the account doesn’t suggest he expected the number to purchase anything.

The failure mode isn’t recording. It’s recording in expectation of a return — which is the transactional trap, and it’s the most common misreading of this whole body of material.

Same ledger. Different purpose. Opposite result.

Why You Need One at All

Briefly, because the argument for tracking is easy to lose in the warnings.

During any long change, results read zero for an extended period. That’s not failure — it’s what a threshold looks like from below. If your only instrument reads output, you’ll get four months of zeros and quit, and quitting will feel like clear-sightedness. That’s the accumulation threshold.

And there’s a second reason, which is more mundane and just as important: unrecorded accumulation disappears.

Forty instances you can’t recall don’t function as evidence. They function as a vague sense that you’ve been trying. A written list of forty functions as a case, and the case is what eventually revises what you believe about yourself.

So the record isn’t only for steering. It’s the material the change is built from.

What This Is Not

It is not an argument for elaborate systems. One line a day in a notebook satisfies every rule above. Anything with dashboards, streaks or scores has already crossed into the other category by design — streaks in particular are scoring devices, since their entire function is to make a broken one feel like a loss.

It is not a claim that feeling good about progress is wrong. Satisfaction is fine. The specific problem is when the record’s function becomes producing that feeling, because then the entry is the payout and the action has been spent.

It is not a substitute for the work. A well-kept ledger of nothing is a well-kept ledger of nothing. This is instrumentation, and instruments don’t move anything.

The Cost

The price is small and unglamorous.

An honest ledger is boring and slightly humiliating.

Boring, because if it’s doing its job it contains one-line factual entries about ordinary actions, with no narrative and no assessment. There’s nothing to enjoy in it.

Humiliating, because rule five means it contains the misses, flatly, with no context. Reading back over a month that includes eleven blank days is unpleasant in a specific way — it’s the absence of any excuse, which you’d normally have supplied automatically.

Second cost: it will show you your actual rate rather than your imagined one, and those differ. Most people are doing considerably less than they think, and the ledger is the thing that tells them.

That’s the value and it’s also why records get abandoned in month two.

Common Questions

How do you track personal change without obsessing over it?
By keeping the record as an instrument rather than a score. Record inputs rather than merit, don’t display a running total, keep it private, review on a fixed interval, and record the misses in the same flat tone as the successes.

When does measuring progress become counterproductive?
When the number becomes a claim about your worth or a balance against which you expect a return. The test is what happens on a bad reading — adjustment means instrumentation, shame or defensiveness means scoring.

Doesn’t counting cancel out the thing you’re counting?
Only when the count is transactional. An action performed to increase a total has collected its return at the point of entry. An action recorded so you can tell whether the practice is happening hasn’t.

Should I share my record for accountability?
No. A record with an audience becomes a performance and gets edited, usually without you noticing. If you need accountability, arrange it separately — a specific commitment to a specific person — and keep the record private.

Why do streaks cause problems?
Because a streak’s function is to make breaking it feel like a loss, which is scoring by design. It also converts a missed day from a non-event into a failure, which is a much heavier thing to recover from.

How often should I look at it?
On a schedule, usually monthly. Frequent checking isn’t reading the instrument — it’s seeking reassurance, and reassurance is a payout.

Where to Start

One line a day. A notebook or a plain file. Nothing with features.

Write what you did, in the fewest words that are still specific. Declined the Thursday request. Twenty minutes on chapter two. Nothing.

That last entry is the one that matters. On the days nothing happened, write nothing, flatly, with no explanation. If you can’t do that without adding context, you’ve found out something about which kind of record you’re keeping.

Then don’t look at it for a month.

At the end of the month, read it once and ask a single question: what does this tell me to adjust?

If the answer is a change — the frequency is wrong, the slot is wrong, the thing is too big — the record is working.

If the answer is a feeling, you’ve built a score, and the fix is in the five rules rather than in trying harder.

I’ve spent twenty-five years working on how ordinary people develop capabilities that look impossible from the outside, and almost all of them kept some kind of record.

None of them could tell you their total.