The machine predicted a strong result for the man in the red coat.
He showed the slip to his companion.
“There.”
“That’s what it gave me for formation work.”
“Perhaps we’re both excellent.”
“You entered endurance.”
Felix moved closer to the booth.
A polished board described a current local assessment followed by a predicted performance category. Beneath it, an attendant invited entrants to place a hand on the marked contact and provide the small trace permitted for the demonstration.
Nobody was being asked to empty a dantian. The reference band was modest. Felix read it twice before paying for his own turn.
The raw pointer moved when he supplied the requested trace. A second indicator settled into a broad prediction category. The attendant wrote both on a slip.
He was embarrassed by how much he wanted the prediction to be flattering. He had qualified yesterday. His Foundation was real. He had perfectly good evidence that a small amount of practiced control could do useful work.
He still glanced at the favorable end of the prediction scale before reading his slip.
The attendant had seen that glance before. He waited with the patient expression of a man selling respectable hopes all afternoon.
Felix reduced his trace within the marked range.
The raw pointer fell.
The prediction did not.
“It can use a broad category,” the attendant said when Felix asked.
“I thought so. Does it use an average?”
“It’s calibrated.”
“From this contact alone?”
The attendant tapped the public diagram.
“Local measurement. That’s the demonstration.”
Felix looked at the separate indicators. A stable prediction wasn’t proof of anything by itself. His own body could produce different small outputs without becoming a different competitor. A device might reasonably use earlier measurements to avoid declaring him three people before lunch.
“May I see the operating note?”
The attendant fetched it from beneath the counter.
By then, the red-coated entrant had returned with his companion. They watched Felix read the part describing the local test range.
“Does it predict arguments?” the companion asked.
“Only if the argument’s already entered in its records,” Felix said.
The attendant didn’t laugh. Felix showed him the relevant line instead of trying again.
“I want to distinguish the current reading from whatever makes the prediction stable. Could we compare the two using your normal check?”
“You can’t open the sealed housing.”
“You operate it. I watch. Nothing beyond your ordinary test.”
The attendant read his own note, then called over the booth’s calibration worker.
She brought the standard reference fitting.
“Two loads,” she said. “Both within the public check. If the pointer doesn’t recognize them, we close the booth.”
Before the worker changed anything, the red-coated entrant offered to repeat his own reading. She explained the permitted range, and he supplied two small traces while Felix watched.
The raw pointer moved by less than it had for Felix. The predicted category stayed still again.
“There,” the man’s companion said.
“Not enough to decide,” Felix replied. “Those two readings might both belong in that category.”
The entrant looked annoyed that his useful demonstration hadn’t immediately become evidence of his excellence or the machine’s dishonesty.
“Then why do it?”
“To find out whether it answers the question. It doesn’t, clearly enough.”
The calibration worker disconnected the ordinary contact and showed them the stamped reference fitting. Its two known inputs would give them a comparison that didn’t depend on guessing how closely a person had repeated himself.
Felix got a blank observation slip.
“And the prediction?”
“That’s what you want to compare.”
The worker set the fitting on the counter before connecting it. Two stamped settings gave an input she could identify without relying on an entrant’s willingness to repeat an effort exactly.
“Same supply,” she said. “Same permitted connection. I change only this setting.”
Felix drew two columns.
“Raw reading. Prediction.”
“And mode,” she said.
He added a third.
The red-coated man’s companion peered over his shoulder.
“Are you examining it or teaching it to write?”
“I’m teaching me to remember what actually changed.”
The worker waited until the ordinary contact was clear. She seated the reference fitting, read the source band and asked Felix to repeat the displayed values aloud while she noted them independently.
They matched.
The first reference gave the expected raw value. The second gave another. The predicted category remained on the same mark.
The attendant spread his hands.
“A broad category.”
“Still possible,” Felix said.
The calibration worker checked the mode lever. The public note showed the check state as local reference only. She set it herself and repeated the comparison.
The raw display changed correctly.
Beside the prediction indicator, a smaller marker remained active.
“What feeds that?” Felix asked.
The worker followed its route around the side of the housing. One line ran from the local contact. Another entered beneath the rear cover and bore a reference tag.
She read it without moving the cover.
“Configured reference.”
“Does local check isolate it?”
She went back to the operating note. This time she read the small supplemental instruction attached to its inner fold.
“Not in this configuration.”
The attendant leaned over her shoulder.
“That wasn’t on the public drawing.”
“No,” Felix said. “It wasn’t.”
He wrote the apparatus identifier at the top of his slip.
The calibration worker pointed out the line’s tagged direction and the state it remained in. She did not claim to know where the value originated beyond the identified reference input. That information wasn’t on the booth note.
“Could the prediction simply be held from the last turn?” he asked.
The worker checked the reset indication. It showed the completed reset required for the standard comparison. She repeated that part herself and watched the reference marker remain active.
“Not an uncleared last turn,” she said. “I can confirm that much.”
Felix crossed out his first possible explanation rather than pretending he hadn’t considered it.
The attendant read over the correction.
“You thought it might be our reset?”
“It was something we could check.”
“We did reset it.”
“Yes. So the note should say you did.”
The man’s shoulders relaxed slightly. Felix knew that movement from work: the moment a person discovered the question concerned the fault rather than their right to be standing beside it.
Felix asked her to mark exactly what she’d checked.
She did.
A contestant at the counter had begun telling the people behind him that the machine was fixing everyone’s result.
“Mine said I’d do poorly,” he announced. “I knew that couldn’t be a real measurement.”
The red-coated man clutched his favorable slip.
“Mine was accurate.”
“Have you done the event?” Felix asked him.
“Not yet.”
“Then keep the slip. We haven’t compared it with a result.”
The disappointed contestant wanted a stronger statement. Felix showed him the two raw observations.
“Those changed correctly. The prediction also uses a reference the public explanation doesn’t describe. We don’t know what that reference is doing yet.”
“But it’s false.”
“The description is incomplete. That’s what we’ve observed.”
A supervisor arrived before the crowd could improve the claim further.
The attendant explained the test and handed over the supplemental note. The calibration worker repeated the configuration state. Felix kept his own hands clear of the apparatus while the supervisor examined it.
“Are you alleging deliberate manipulation?”
“I’m reporting an extra input. I can’t identify its purpose from this counter.”
“It may be an authorized baseline.”
“Then the explanation should say the prediction combines it with the local measurement.”
The supervisor read the public board.
For a moment, he looked as though he would argue with the wording purely because Felix had noticed it first.
Then he asked the attendant for a fresh notice card.
“Prediction demonstration under reference review,” he dictated. “Current local readings remain separately identified.”
The supervisor made the disappointed contestant repeat the claim he’d heard. It had already become a story about purchased victories by the time it reached the man’s mouth.
“Who said that?” the supervisor asked.
The contestant pointed at Felix.
“I said an additional reference was active during the local comparison,” Felix said.
“That’s not as interesting,” the contestant replied.
“It’s what I can sign.”
The supervisor wrote that sentence into the observation instead of the larger accusation. Felix checked the wording before agreeing to it. Being less interesting was unexpectedly difficult when a crowd had assembled to find him useful.
The calibration worker restored the ordinary setting and checked the raw scale. The supervisor recorded the observed configuration before authorizing any further change to the reference arrangement.
Felix asked for a copy of the operator note with the apparatus identifier and comparison conditions.
“Not the device?” the attendant said.
“It would make my pocket worse.”
The man looked at the slate at Felix’s belt and almost smiled.
The signed observation would be ready for collection the next morning. The supervisor wanted the calibration worker’s final check attached before releasing it.
Felix agreed. He had the observations in his own hand, witnessed at the booth, and a collection time.
He walked away pleased.
The pleasure was familiar. A claim small enough to test. An unexpected path. Someone willing to check it instead of insisting the diagram had to be right.
At home that evening, he copied the results for Lyra and Barnaby. She added the limited shipment sequence she’d obtained at the reception, keeping it on a separate page.
“The representative knows we’re asking about transport now,” she said.
“The booth supervisor knows I’m interested in an omitted reference.”
“Neither fact tells us the source.”
“No. But they’re both better questions.”
Barnaby checked the scheduled collection time against the next day’s practice.
“You have enough time if the collection remains a collection.”
“It’s a signature and a page.”
“A phrase frequently preceding an additional hour.”
Felix promised to be at the lane before Jian’s paired exercise.
The next morning, the signed note lay ready at the booth.
It should have ended there.
Felix read the apparatus identifier, the two reference values and the supplemental instruction number. One word in the copied configuration line was ambiguous.
“Does this mean supplied baseline or retained baseline?”
The attendant called the calibration worker.
She checked the original.
“Configured reference. The abbreviation was copied badly.”
Felix thanked her and marked the correction.
Then he asked whether they’d recorded the time of the normal-state restoration.
They had. The worker fetched the log to add it.
Five minutes became a useful clarification. Another five became a discussion of what the device’s public notice now claimed. Felix could feel each question fitting beside the previous one, forming a better account.
He could also see the practice lane clock from the booth’s side opening.
A pair of entrants passed carrying the spring return frame Jian had booked. Felix recognized the covered brass end and suddenly remembered his own marked place beside it.
He hadn’t forgotten practice existed. He’d kept assuming the next question would fit inside the remaining time, even after he stopped knowing how much remained.
The lane clock was entirely indifferent to that distinction.
He looked at it too late.
The paired exercise had already begun without him.
Or tried to.
Felix closed his mouth around the next question.
“I need to go.”
The attendant handed him the finished note.
“You’ve got the conditions now.”
He did. Device identifier. Local loads. Actual raw changes. Stable prediction. Active additional reference. No proof yet of its originating value or purpose.
The signature belonged to the supervisor. The calibration worker had initialed the correction beside her own observed state. Felix hadn’t needed to take either person’s word for a result they couldn’t show; they’d put the comparison where another reader could follow it.
He wanted to bring them a good answer later.
An excellent, genuinely useful small result.
He tucked it into his bag and hurried toward the training lane, rehearsing how quickly he could explain why it had mattered.
Halfway there, he realized the explanation contained no part where Jian had agreed to wait.