12. After the Ninety Days — Building Your Own Drill and Knowing When to Run Nothing
Before you read another paragraph, go find the recording from Baseline Day. The one you made in Chapter One, before you knew anything, on the day you were told to make it and not listen to it again. It has been sitting untouched for ninety days for exactly this reason. Play it. Score it with the same sheet you used then — challenges attempted, challenges landed, mean latency from the end of their utterance to the start of yours, and the count of moments where you now hear an opening you did not hear at all the first time.
Do this before you keep reading. Not because there is a trick waiting in the rest of the chapter that would contaminate the score, but because reading the chapter first will give you a story about your progress, and the story will be more flattering and less useful than the number. The number is the only thing here that does not care how you feel about your development.
When you have scored it, three things are typically true, and it is worth knowing in advance which of the three you are looking at.
The first: mean latency has dropped, usually by somewhere between a third and two-thirds, and the drop is not uniform. Look at where it fell most. In almost every log I have seen, the meta-model challenges from Week Four fall hardest — they are the most heavily drilled and the most mechanically retrievable. Reframing from Week Eight often falls least, because selection is a judgment and judgment does not compress the way retrieval does. That unevenness is information, not failure. It tells you which capabilities are now procedural and which are still running through deliberate search.
The second: your count of heard openings has gone up more than your count of landed challenges. Intake outruns production. This is the expected shape and it is worth naming plainly, because it feels like a deficit and it is actually the correct order of development. Weeks One through Four built the ear before the mouth deliberately. You will hear four things you could have done for every one you did. That ratio narrows with use and never closes, and practitioners who tell you it has closed for them are, without exception, no longer hearing everything.
The third, and the one nobody warns you about: some of your Baseline Day conversation now reads as fine. You will find stretches where you ran nothing and, listening back with ninety days of ear, you can see that running nothing was right. Mark those. Count them. That count is the subject of the second half of this chapter and it is the number that will matter most to you in a year.
The method, extracted
What you actually bought with the ninety days is not twelve patterns. Twelve patterns is a small return on that much attention, and you could have got most of them, badly, from the book you already read. What you bought is the method underneath the weeks — the thing that let a described capability be turned into a graded, scorable, repeatable unit of work. That method is transferable to every capability anyone will ever describe to you, in this discipline or any other, and it is five requirements long.
An isolatable unit. The capability has to be cut small enough that one repetition takes seconds and produces one bit of information. Week Three isolated classification from intervention and the entire week's value came from that cut: sorting a sentence into deletion, distortion, or generalisation is one unit; challenging it is a different one. When people fail to build working drills, this is where they fail — they drill "rapport," which is not a unit, or "being present," which is not even a capability, it is a description of a result. Ask of any candidate: can I do one of these in under ten seconds, and will I know afterwards whether that one was right? If no, cut smaller. Keep cutting until yes.
A stimulus you did not author. This is the requirement most self-designed practice violates, and it is fatal when violated. If you write the sentence you then challenge, you have written a sentence your existing competence can already handle — your generation and your recognition come from the same store, so the drill tests nothing and confirms everything. Every week in this book pulled material from outside your head: transcripts, recordings, podcasts, a partner, strangers in a queue. The stimulus must be able to surprise you. A drill that cannot surprise you is a rehearsal, and rehearsal builds fluency at exactly the material you rehearsed and nothing adjacent to it.
A stated latency. Not "quickly." A number, decided before you start, enforced by something that is not your judgment — a timer, a metronome, a partner counting. The number is the drill's difficulty dial and it is the only one you should turn. Week Four ran at four seconds and dropped to two by Friday; the drop, not the absolute value, was the training. Set it where you succeed roughly seven times in ten. Above nine in ten and you are grooving something you already own. Below five in ten and you are practising the sensation of failing, which consolidates about as well as you would expect.
An external check. Something outside your own sense of how it went decides pass or fail. This is the requirement that makes the difference between a drill and a habit. A partner who scores you, a recording you re-listen to cold, a written key, a physiological sign you agreed on in advance. In Week Six the check was the anchor firing on a cold test after a break — a condition you cannot talk yourself past, which is precisely why it was chosen. Your in-the-moment assessment of your own performance is systematically generous and systematically fastest to inflate at exactly the skill level you are at now.
A failure mode written down in advance. Before the first rep, write the sentence: this drill goes wrong when ___. Naming it in advance is what lets you catch it, because in the middle of a week of reps you will not notice drift; you will only notice that the reps feel good. Week Seven's written failure mode was matching becomes mimicry and the other person feels handled. Week Nine's was interrogation becomes badgering and the client starts performing outcomes to end the questioning. Every drill in this book has one because every drill in this book has an edge past which it produces the opposite of the thing it trains.
Five requirements. Miss the stimulus one and you build confidence without competence. Miss the check and you build a story. Miss the written failure mode and you build the shadow of the skill and cannot see that you have.
Building one in front of you
Take a capability this book never taught: holding a silence after someone finishes speaking without filling it, long enough that they continue into the material underneath. It is a real capability, it is described everywhere in the therapeutic literature and drilled almost nowhere, and it is a good test case because it looks undrillable. Watch the five requirements do the work.
Isolatable unit. One rep is: person stops talking; I do not speak; I count; they either continue or the silence breaks. Under ten seconds is doubtful — the whole point is the duration — so call it one rep per opportunity, up to twenty seconds. Binary outcome available: did I hold to my stated count, yes or no. That is a unit.
Stimulus I did not author. Live conversation only. This one cannot be drilled on transcripts, which is a real constraint and worth stating rather than fudging: the drill requires a partner or an unwitting public. I take the unwitting public — three ordinary conversations a day, one target rep in each, nobody told. The rep is not the whole conversation; it is one selected moment inside it.
Stated latency. Inverted here, which is the interesting part. Every previous drill made the number go down; this one makes it go up. Days one and two, three seconds. Days three and four, five seconds. Days five through seven, eight seconds. Eight seconds of silence after someone finishes a sentence is very long and will feel to you like a social emergency. That feeling is the load the drill is training against, the way the shot clock was the load in Week Four.
External check. Two, because self-report on silence is worthless — your sense of elapsed time distorts under exactly the discomfort this drill induces. First: a watch with a second hand, glanced at, or a silent count you calibrate each morning against a timer so your internal three is a real three. Second, and better: did they continue speaking? Continuation is external, observable, and not fudgeable. Score each rep on two columns — held to count and they continued. Continuation without holding is luck. Holding without continuation is a legitimate pass on the drill and a data point about that person.
Failure mode, written in advance. This drill goes wrong when the silence stops being spacious and becomes a withholding the other person can feel. There is a version of silence that invites and a version that pressures, and the difference is not duration — it is what your face and body are doing while you wait. If people start filling silence anxiously rather than continuing thoughtfully, if they say "sorry, am I going on too long," you have crossed the edge and you are running a pressure technique, not an attention. The correction is not a shorter count; it is softening the face and letting the body settle while holding the same count.
Scoring sheet, which is three columns and a line: date, rep number, count target, held (Y/N), continued (Y/N), and one field for what they said in the extra space if they used it. Success condition for the week, stated before starting: by day seven, twelve of the last fifteen reps held to an eight-second count, and at least eight of those twelve produced continuation. That is a number I can fail. Write yours so you can fail it too. A success condition you cannot fail is a wish.
Notice what just happened: a capability that looked like temperament — patience, presence, whatever the seminars call it — decomposed into a countable unit with a scorable outcome and a written edge. Almost all of them do. The ones that resist are usually not capabilities at all but descriptions of a state that results from several capabilities, and the move is to ask what someone with that state visibly does that someone without it does not, then drill the doing.
Maintenance
Skills decay at different rates and the differences are not subtle, so a maintenance schedule that treats them equally wastes most of its time.
Fastest to go, in every log I have kept and every one I have been shown: latency on the meta-model challenges. Not the knowledge — you will know a lost performative for the rest of your life — but the speed of retrieval, which slides back toward deliberate search within about three weeks of disuse. Second fastest: calibration, from Week Five, which does not decay so much as get overwritten. Interpretation creeps back in and reoccupies the ground sensory description held, and it does this invisibly because interpretations feel like observations. Third: anchoring, which decays in the setting rather than the firing — your anchors keep working, but your ability to set a clean one under live conditions degrades.
Slowest to go, and effectively permanent once built: transcription-grade hearing. Once you can hear surface structure, you cannot stop hearing it. Also durable: the six reframes from Week Eight, which live in retrieval rather than production and behave like vocabulary. And the refusal from Week Eleven, which is the reassuring finding — an ethic drilled to latency stays drilled, provided you keep meeting the situations that call on it.
So the weekly block, fifteen minutes, and it is fifteen because a maintenance block you will actually run beats an hour you will run twice and abandon.
Five minutes on challenge latency: any podcast, any interview, shot clock at two seconds, challenge aloud into the room. Do not evaluate quality. Speed only. Five minutes on calibration: one person, in life or on video with the sound off, and write only what you saw — muscle, colour, breath, rate — stopping the moment a conclusion appears in the writing, then continuing beneath it. Five minutes on whatever your Day 90 retest showed as your worst column, rotated as that column changes.
Put it in the same slot each week. The specific slot matters more than which slot, for the ordinary reason that the environment cues the behaviour and a floating practice has no cue.
Your turn
Design a seven-day drill for a capability of your own choosing, using the five requirements. Choose something this book did not teach and that you actually want — the strategy elicitation you keep half-doing, the ability to hold your own state when someone escalates, the sub-modality shift you have read about for years and never run cleanly. Write the brief on one page: unit, stimulus source, latency and its progression, check, failure mode, success condition with a number in it.
Then run it. Seven days, scored daily, no adjusting the success condition after day one — if it turns out to be too hard, you fail this week and reset the number for next week, which is a completely acceptable outcome and the only one that keeps the number honest.
Then, and this is the part that most people skip and should not: give the brief to a partner and ask them to break the success condition. Not to encourage you. To find the way you could satisfy your own criterion while not having the skill. Give them the explicit instruction: find how I could pass this without being able to do it. They will find something, usually within a minute, and it is usually the check — checks are where self-designed drills go soft, because we write the check while imagining ourselves succeeding at it. Fix what they find and run the week again with the repaired condition. A drill that has survived a hostile reading is worth ten you wrote alone.
The turn toward silence
Go back through your logs — all eleven weeks — and count the entries where you wrote some version of saw the opening, didn't take it, and that was right.
In the first four weeks you will find almost none. There is a reason for that beyond inexperience: in the first four weeks you could not see many openings, and the ones you saw you took, because taking them was the assignment and because a newly available move exerts a pressure to be used. Somewhere around Week Seven or Eight the entries start. By Week Ten and Eleven, in most logs, they are the majority of the interesting entries — the moments the practitioner found worth writing down are increasingly the moments they declined.
That trend is not a slackening. It is the thing the ninety days were manufacturing, and it took until now to say so plainly.
Here is what the drills were actually doing. Every repetition trained two things at once and you were only ever scored on one of them. Scored: production — the challenge, the reframe, the anchor, the recovery. Unscored, and accumulating the whole time underneath: perception. Several hundred forced repetitions against real material did not merely install responses; they rebuilt what your attention automatically parses, so that surface structure, incongruence, state shift, and the shape of someone's model now arrive already sorted whether or not you intend to do anything about them. That is why the openings-heard number climbed faster than the challenges-landed number. Intake was the larger product all along. Production was the forcing function that built it.
And once perception is automatic, the discipline inverts. When you could not see the openings, competence meant learning to make moves. When you cannot stop seeing them, competence means selection — and selection, in a life mostly made of conversations that are already going fine, resolves overwhelmingly toward not intervening. Most of what you now perceive does not need you. A friend generalising about their family over dinner is not presenting a lost performative for challenge; they are being a person at dinner, and the generalisation is doing work — holding a complicated history in a shape that lets it be spoken about at all. Your challenge would be technically correct and would take something from them that they were using.
This is the turn, and it is worth stating without softening: the drills do not build a set of moves you deploy, they build a perception you cannot switch off, and the final competence is not adding the pattern but declining to interrupt what is already working. What began as a training in production ends as a training in restraint. The noticing that remains when you stop intervening — that is the thing the ninety days were actually for. The patterns were scaffolding. Scaffolding is not a failure to be a building; it is how the building got made, and it comes down.
Which is not a licence to run nothing forever. A practitioner who never intervenes and calls it wisdom is usually just avoidant, and the distinction between restraint and avoidance is empirical, not philosophical: could you have run the pattern cleanly at that moment, at latency, if you had chosen to? If yes, declining is restraint. If no, declining is a story you are telling about a capability you have let go. Week Eleven's refusal was drilled precisely because an unpractised ethic loses to a practised technique; the same logic runs in reverse here. An unpractised technique makes your restraint meaningless, because you were never going to do it anyway.
The failure mode of the whole book
Now the edge, and it belongs to the entire ninety days rather than any single week.
The characteristic casualty of this training is the practitioner who has run every drill and can no longer have an unpatterned conversation. They are not a manipulator — that is the failure everyone anticipates and it is comparatively rare, because it requires a coldness the drills do not create. The real casualty is warmer and sadder. They are permanently in intake. Their partner says something about their mother and they hear a nominalisation, a comparative deletion, and a shift in breathing at the third clause, and somewhere in that parse the person disappeared and a structure took their place. They have not stopped loving anyone. They have stopped being reached. Being reached requires a moment of taking the utterance at face value, at its own valuation, without the sorting — and the sorting is now automatic, which was the entire point, which is why this failure mode is not a defect in the training but its own shadow.
You will know it by a specific symptom: conversations feel like work when nothing is being worked on. Also by a report from someone close to you that they feel studied, which is a word that comes up so often it may as well be diagnostic.
The practice that returns you is narrower than you would guess and it is not "stop analysing," which does not work, because you cannot stop a perception by deciding to. It is this: in one conversation a day, with someone you are not helping, deliberately give your attention to content — to what is being described, to the mother, the job, the fear, as things in the world rather than as evidence of a model. When the sorting starts, and it will, let it run and do not act on it, and put your attention back on the thing they are talking about. The sorting continues in the background. It has to; you built it. But it is running behind you rather than between you, and that positional difference is everything.
Do this daily for a fortnight and the studied feeling recedes, and — this is the part that surprises people — your pattern work gets better, because material offered by someone who feels met is richer than material offered by someone who feels processed. The restoration is not a sacrifice of skill. It is a condition of it.
The closing rep
One hour, this week, with someone you love. Not a client, not a partner from the drills, not a colleague you are secretly practising on. Someone whose good opinion you would miss.
Run nothing. No challenge, no reframe, not the mildest bit of matching that you would swear is just being a decent listener. If they say everyone thinks they are difficult, let it stand — the sentence is not a symptom, it is something they wanted you to know. If they generalise, agree or disagree the way you would have at twenty. Take the hour at face value.
You will fail at this several times inside the hour. The challenge will be halfway out of your mouth before you catch it, which is exactly what four hundred milliseconds of procedural memory was built to do and is not a reason for guilt. Catch it, let it go, come back to what they were saying.
Afterwards — not during, not with your phone on the table, but afterwards, sitting somewhere by yourself — write down what you noticed anyway. Everything the perception delivered while you were not acting on it. The place their voice changed. The thing they said twice. The subject that arrived three times from three directions and was never named. The moment they were about to say something and did not.
That page is the last artefact of the ninety days, and it will be the most detailed reading of another person you have ever produced, made in an hour where you ran nothing at all. Keep it. When you next wonder whether the drills worked, that page is the answer, and it is a better answer than the latency column — which was only ever the measure you needed while you were still becoming able to do the thing you can now do without doing anything.
Practice — Chapter 12
The sentence you are converting will almost never arrive as a drill. It arrives as praise for someone else — a line in a review, a remark a coach made once, a description in a book of what the good ones do. Praise is a description of an outcome. A drill is a description of a rep. The whole method is the distance between those two things, walked deliberately, five steps at a time.
The steps: find the verb. Find the load. Build the ladder. Write the success condition. Set the count and the stop.