The commercial course promised fluency through volume. Its brochure showed rows of trainees moving in synchrony beneath the phrase TEN THOUSAND CORRECT REPETITIONS. Elias had liked the number because it converted uncertainty into inventory. Perform enough work and technique would exist at the end like a completed order.
Selma crossed out the slogan on his printed schedule.
“A repetition is not correct because you intended to improve,” she said. “And bodies do not accept bulk delivery.”
Elias corrected his rear foot and ruined his shoulders.
He lowered his shoulders and lost the spear line.
He restored the line and forgot to breathe.
Selma tapped the stopwatch.
“Again.”
The first coached block established a baseline slow enough to be humiliating. Selma removed the target, the timer, and the spear. Elias stood between two tape lines and transferred weight from rear foot to front until the heel, knee, and hip travelled without crossing them.
At natural speed, the rear foot turned outward on twelve of twenty repetitions. When he slowed to four seconds forward and four back, it turned on five. When Selma shortened the distance by ten centimetres, it turned on one.
“That is your present range,” she said.
“The exam requires more.”
“The exam requires movement you can control. Range grows after control.”
He wanted to practise the longest stance because it resembled the final task. Selma made him practise the short stance because it produced a pattern worth reinforcing.
“Which correction?”
“One.”
“There are three wrong.”
“There are more than three wrong. Choose one.”
They were six weeks into training.
Weeks one and two contained no full-speed thrusts. Ankle mobility, hip strength, casualty-harness starts, and breath under low simulated pressure. Elias's log showed progress too small for anyone watching once a week to notice. Right ankle reach moved from seven centimetres to seven and a half. Fifty transfers ended with fewer heel lifts. His sleep average rose because Selma cancelled training when it did not.
In week three she returned the spear without a target. In week four she added the target without time. Week five added one peripheral light. Week six allowed the list to become visible.
Elias’s log contained hundreds of slow thrusts, weight transfers, recovery steps, and casualty drags. Analysis had become eager enough to make the work worse.
Rear foot rotation: 14° outside baseline. Hip rotation begins early. Forward shoulder elevation: high. Recovery width: narrow.
Every movement arrived carrying a list.
Elias tried to fix the list.
His body tried to move.
Neither succeeded.
That was the second anchor session: every variable at once.
Selma gave him a simple instruction—touch the centre pad and recover behind the tape. Elias activated Analysis before moving. Rear foot fourteen degrees out. Hip early. Shoulder high. Recovery narrow. He corrected the foot and delayed the hip. The shoulder rose to compensate. He lowered it and lost the straight line to the target. He found the line, forgot the recovery, and ended with both feet crossed.
The first repetition took eleven seconds.
The second took fourteen because he approved each joint before allowing the next one to move.
On the third, Selma tapped his helmet from the side while the spear remained extended. Elias could not recover without unthreading his internal checklist.
Selma covered the floor mirror with a curtain.
She also disabled Analysis display for the remainder of the block. Elias could use the skill to record comparison afterward, not narrate the motion while it happened.
Without the display, error felt larger. Elias could sense that something was wrong and could not name it fast enough to fix it. The urge to reopen Analysis arrived before each movement like an itch beneath armour.
Selma gave him an external cue. A strip of tape under the rear heel. If the shoe crossed it, stop and reset. No angle, no diagnostic list, no argument about why.
On the fifth repetition, his shoulders lifted so far the spear point missed the pad. Elias stopped.
“Did the heel cross?” Selma asked.
“No.”
“Then complete the recovery.”
“The strike failed.”
“The drill did not.”
He recovered with an ugly shoulder line and a stable rear foot. The distinction was almost physically painful.
“Today, rear foot only,” she said. “Shoulders may offend you.”
“The habit will reinforce.”
“So will paralysis if you need to consciously approve every joint.”
She placed tape around his rear shoe. Elias transferred weight forward without crossing the line.
Slowly.
Again.
The shoulders offended him. The spear point wandered outside the centre pad. Selma let both happen. At the end of ten repetitions, they reviewed only the rear-foot video. Seven stayed inside the tape. On the next set, eight.
“Now shoulder only?” Elias asked.
“Tomorrow. Your nervous system is allowed to finish one lesson.”
Again.
At repetition thirty, his ankle tightened. At forty-two, the heel lifted. Selma stopped the block at fifty and made him record the failure curve.
He charted repetitions one through fifty against heel angle. The error remained low until thirty-four, rose gradually, then jumped after forty-one. Fatigue was not a general excuse attached after bad work. It had an observable onset.
Selma set future blocks at thirty clean transfers followed by rest, not fifty performed for character.
Workdays complicated every baseline.
Selma refused to schedule around an imaginary rested Elias. Once every two weeks, a coached session followed an ordinary workday by design. The goal was not maximal performance. It was learning what condition arrived at the hall and reducing the task before fatigue selected the lesson.
They began tagging sessions by preceding load. Surface analytics meant stiff hips and rested grip. Hachinger harvest meant calf fatigue, shoulder load, and earlier form loss. Clinic night meant interrupted sleep even if Elias had done no physical work. Identical drills no longer pretended to be identical samples.
The tags revealed patterns Elias had called inconsistency. His right ankle lost range after long periods in interior boots. His grip tired earlier after Sillglass paddles than after Hachinger frames. Clinic nights produced accurate first sets and unusually poor attention recovery, even when sleep duration met the minimum because worry had broken it into pieces.
Selma adjusted the programme. Mobility before removing work boots. No heavy grip training after Sillglass. Attention drills moved away from clinic mornings. None of the changes increased weekly volume. Performance improved because the same volume stopped colliding with predictable fatigue.
Daniel’s footage added another problem.
Daniel attended the third anchor in person and wore standard loan armour instead of his fitted response suit. Even constrained by unfamiliar equipment, he moved with a surplus Elias could see. The armour changed Daniel's reach. It did not erase Level Twenty.
The third anchor session used side-by-side video. Selma aligned Daniel's Level-Seventeen shield-lane thrust with Elias's best slow repetition. At first, Analysis made Daniel's movement the reference and reported Elias late at every phase.
Selma changed the question.
“Which joint relationships remain stable in both?”
The point stayed behind the front knee until commitment. The rear heel transmitted force without collapsing inward. The shoulder rose only after the target line was complete. Recovery began with the threat-side foot, though Daniel reclaimed ground and Elias yielded it.
“Those are transferable,” Elias said.
“Now show me what is not.”
Daniel's stance length. Front-leg drive. Acceleration. The distance covered during recovery. His ability to remain upright after committing more body weight than Elias could stop.
At Level Twenty, Daniel accelerated through motions Elias could barely segment. He recovered from an extended thrust by driving off the front leg with enough force to reposition his entire body. Analysis labelled the technique efficient.
Elias tried it.
His knee collapsed inward.
Selma caught the shaft, not his body. Elias recovered to the mat without impact and logged the knee line as a capacity failure, not a technical mystery.
Selma paused the footage.
“Transferable?” she asked.
“Front-leg drive.”
“At his Power and Agility.”
“So I scale it.”
“Or choose a recovery that does not require attributes you lack.”
She showed an older rescue clip of herself at Level Nine. Her stance was shorter. The recovery yielded ground instead of reclaiming it. Less impressive. Stable under load.
In Selma's clip, a shield specialist occupied the ground she yielded. Her technique was not universally safer; it assumed a team geometry where retreat opened space. Alone, giving ground could surrender the casualty. Conditions lived inside every model.
“Margins,” she said.
Elias adopted the shorter line.
Daniel held the casualty lane while Elias tested it. On the first repetition Elias yielded too far and stepped outside the shield edge.
“I can widen,” Daniel said.
Selma answered before Elias. “Not yet. He learns the standard lane before a stronger teammate rescues the geometry.”
They reset. Elias shortened the yield until his shoulder remained behind the shield and his front foot cleared the dummy harness. It sacrificed another fraction of reach. It preserved the team.
Afterward, Daniel tried Selma's shorter recovery. His body completed it easily and left him with unused force.
“Feels inefficient,” he said.
“For you, often,” Selma replied. “That does not make your efficient version available to him.”
Daniel nodded. No encouragement, no denial, no story about wanting it enough.
His times improved by fractions. By week eight, his best full route was fifty-seven seconds and his median sixty. Selma tracked the median. Elias kept looking at the best.
“The examiner gets one candidate,” she said. “Not a highlight reel.”
By March, fifty slow transfers became one hundred. Threat drills became less disastrous. He still missed the standard.
“Less disastrous” had a measurement. Selma placed four lamps around the casualty lane and activated one while Elias fitted the harness. In February, he either missed the cue or abandoned the buckle to look directly at it. By March, he could call colour and position while keeping one hand on the casualty.
Then she changed what the lamps meant. Red no longer always closed the route. Sometimes it marked a contained threat requiring only acknowledgment. Green sometimes opened a path and sometimes indicated a target outside Elias's role. He had to hear Klara's recorded instruction, retain the current objective, and decide whether the cue altered it.
His first score fell. Pattern recognition had learned the colours faster than judgment learned the conditions.
They stripped the drill back again. Elias stated three things before movement: casualty destination, command source, current stop condition. The words cost one second at the start and saved several after the cue.
At the end of the month, his best full route reached fifty-five seconds. The median remained fifty-eight. Passing required fifty-two.
Six seconds was small enough to imagine stealing and large enough that every attempted shortcut showed somewhere else: a twisted strap, a missed lamp, a breath held too long, a knee asked to stabilise after fatigue had removed the capacity.
Analysis climbed through applied self-comparison without reaching another level.
The training produced no integrated experience. The municipal hall was controlled, the patterns repeated, and no live objective depended on the outcomes. Elias gained strength by ordinary adaptation, technique by coached correction, and records by effort. The System did not need to reward every useful process for it to matter.
Some evenings that principle felt mature. On others, he opened his status after two paid hours and resented the unchanged number.
One Friday, Elias arrived after six interior hours and insisted on completing the scheduled session.
The fourth anchor began before he picked up the spear. His pain log read zero. Fatigue read seven. Sleep had met minimum. He had eaten and hydrated. No rule automatically cancelled the block, which made judgment necessary.
Selma tested ankle range. One centimetre below morning baseline. She tested single-leg hold. Right side failed at twenty-two seconds instead of the usual thirty. She reduced the planned eighty movements to forty and removed loaded recovery.
“Fatigue seven,” Selma read from his log.
“The exam won’t wait for me to feel rested.”
“The exam also does not follow a work shift.”
“Real incidents do.”
Selma looked at him. “Reduced block. Stop at form loss.”
They defined form loss before starting: two consecutive heel departures, knee crossing the inner tape once, or shoulder rise preceding hip rotation twice. Elias repeated the criteria. No one would decide afterward that a nearly perfect movement deserved continuation.
They began.
The first twenty thrusts held. At thirty, his shoulder rose. At forty-one, the rear heel drifted outward.
“Stop,” Selma said at forty-one.
“One deviation.”
“Block was forty.”
Elias had counted the warm-up separately. Selma had not. He was already performing an extra repetition while arguing about whether it existed.
Analysis displayed the deviation.
Elias corrected.
The next thrust aligned perfectly.
It was repetition forty-two.
Hip, knee, and foot transferred force in one line. The spear point stopped inside a circle smaller than a coin.
Hip rotation: within stable reference. Rear-foot alignment: within stable reference. Force-path deviation: minimal.
His support leg buckled.
The force path demanded strength his fatigued hip no longer supplied. Mechanical alignment removed the compensations that had previously shortened the motion. For one clean instant, the full load travelled through tissue at its limit.
Elias struck the mat on one knee. The spear clattered away.
Selma was beside him before the System box faded.
“Pain?”
“No. Leg gave out.”
“Sit.”
“The form was right.”
“Yes.”
“Analysis says—”
“I can read.”
She tested the knee, ankle, and hip. No injury. Fatigue had removed the strength required to support the corrected line.
“The motion was mechanically better than your body could currently sustain,” Selma said. “That is not success under field conditions.”
She showed him the trace. Alignment entered the stable band. Support force exceeded his fatigued capacity a fraction later. Analysis had celebrated the first fact before the second completed.
Elias sat on the mat staring at the spear.
“How many repetitions until it is?”
“Enough correct ones. Enough strength. Enough recovery. Not a sacred number.”
“Ten thousand swings.”
“If you perform ten thousand bad swings, you become very reliable at failing.”
She ended the session.
Nika reviewed the force trace remotely because the household plan required a medical check after any collapse. She found no acute injury, imposed forty-eight hours without loaded movement, and Selma cancelled the next coached block. The medical clause waived the cancellation fee, but not hall insurance or transit. Recovery occupied the work because work had created the need.
The following morning, Elias could walk stairs without pain. His hip ached like ordinary fatigue; Medical recorded a capacity event rather than injury. The clean report did not make the extra repetition acceptable.
At breakfast, Anna asked what happened. He gave the sequence without waiting for the shared log to tell her: reduced block, count disputed, stop called, extra repetition, leg buckle, no injury, forty-eight-hour restriction.
“Did you hide anything?” she asked.
“No.”
“Did you follow the stop?”
“One repetition late.”
She nodded, unhappy and informed. Disclosure did not turn a mistake into virtue. It kept the mistake from acquiring secrecy as a second injury.
At home, Elias opened the training log. Beside the System’s nearly ideal measurements, he added two fields of his own.
CAPABLE UNDER LOAD: NO
REPEATABLE UNDER FATIGUE: NO
Analysis could tell him a movement was correct.
It could not make him capable of surviving it.