The worked example

Here is the source sentence, taken as it came: Strong operators protect their capacity. They don't say yes in the room.

Step 1 — find the verb. "Protect their capacity" is a result; nothing in it can be done. "Don't say yes in the room" is closer, but it is a negation, and a body cannot perform a not. Every drill built on a negation is a drill with no rep in it. So ask what goes into the space where the yes would have gone. The answer has to be filmable: a specific sentence, said out loud, within about two seconds of being asked. "Let me look at the week and come back to you before end of day." That is a verb. Someone standing in the room could score it without knowing what you were working on.

Step 2 — find the load. What makes that sentence hard is not the sentence. It is who is asking, how much they want it, and how much you want them to think well of you. So the load is the social cost of the pause — and here is the part most people get wrong. Load the mechanism, not the topic. If your over-commitment runs on enthusiasm, then loading the drill with a request you resent trains nothing at all; you were never going to over-commit to that. The hard rep is the ask you would enjoy saying yes to.

Step 3 — build the ladder. Three rungs, and only the load moves between them. Rung one: asynchronous, low stakes, someone junior, you have time to type. Rung two: live, one-to-one, a peer, a real ask. Rung three: live, in a group, someone whose regard you want, asking for something you'd like to do.

Start at rung one even though rung one is easy. You start where you can already succeed so that the first thing you learn is the shape of the rep rather than the taste of failing. But do not stay there, because a drill that feels smooth is under-loaded — this is Robert Bjork's distinction between performance and learning, and it is one of the more reliable findings in the area: how well you do during practice is a poor guide to what you keep, and conditions that depress performance in the moment often improve retention. The mechanism has a ceiling as well as a floor. Error has to arrive at a rate you can survive. Too little and there is no signal to learn from; too much and attention goes to managing your own arousal, and what you rehearse is the threat rather than the skill.

Step 4 — write the success condition. Before the first rep, not after. Here is a first draft that fails: I'll feel less pressure to commit. It fails three ways at once — it is judged by mood, it cannot be falsified at the level of a single rep, and it rewards the wrong thing, since you could feel perfectly calm and still say yes.

Second draft: I said the holding sentence inside the window. Binary, external, scoreable by a bystander. Better — and still not enough, because you can say the sentence and then send the yes forty minutes later from your phone. So pair it. The rep passes when (a) the sentence was said in the window and (b) the answer you sent afterward was the answer you would have given at your desk with the calendar open, which is checkable because you wrote the desk-answer down before you replied, or you didn't.

The general shape: one condition for the behavior, one for the thing the behavior was supposed to protect. The success condition has to survive being gamed by the person it is grading, and that person is you.

Step 5 — set the count and the stop. Twelve reps at rung one, or eight consecutive passes, whichever comes first. But notice that you cannot schedule these — the ask has to come from outside you. A drill built on reactive material is counted in occasions, not calendar slots, which means it needs a trigger you can notice rather than a reminder you can snooze.

Then the stop, which is part of the design and not an afterthought. Retire the drill when, across ten consecutive asks, you did the right thing before you remembered the sentence. Not when it feels easy; easy arrives early and lies. And distinguish retiring from abandoning: if you are still failing rung one after a dozen occasions, the verb is wrong, and the repair is upstream at step one rather than in more repetitions.

Look at what the drill did to those twelve occasions. It turned a live human being's request into a scoring event. That is an acceptable price for twelve occasions and an unacceptable way to be in rooms — which is exactly why you write the exit while you are still designing, because the person who decides when to stop will be a person with months invested in the running.

The objection worth taking. The strong claims made for deliberate practice have not held up well; the meta-analytic work found that accumulated practice explains a minority of the variance in performance between people, and much less of it in professional work than in chess or music. Fair. But that is a question about the gap between the best and the rest, and it is not the question here. The claim this method makes is narrower and better supported: a behavior you can specify, observe, and load will move. What the evidence should make you doubt is the transfer from moved behavior to excellence — and note where the failure concentrates. Practice does worst in exactly the domains where nobody can say what good looks like. That is a constraint on step one, not a defeat. If you cannot write the verb, do not expect the reps to rescue you.