He used them.

  Not all at once. Not dramatically. The way you used time when you understood that the measure of ten days was not how much you could force into them but how precisely you could direct each one toward what actually needed to happen.

  Day one: Zone Three, morning. Efficiency drill across all three differential points, modified to test interaction effects under sequential loading. Notebook entry: the pockets were independent. Loading in one didn't affect saturation in the others. That meant he could design a training circuit that moved between them at intervals calibrated to his system's recovery cycle—maximize total exposure to elevated saturation without overloading any single session.

  Day two: intermediate library texts, afternoon. The calibration study's original reservations were becoming clearer the more he read around them. The study had flagged three edge cases where the dissipation assumption broke down under specific field conditions. The cases had been classified as theoretical: possible in principle, not observed in practice. He wrote two pages of notes and flagged a citation that pointed to an earlier instrument design document he hadn't found yet.

  Day three: training hall, fifth bell. Standard load test—sixty through eighty-five percent, held at each level, reading the external output numbers against his internal sense of his system. He was building a comparison map: what the instruments said at each load level, what he felt at each load level, where the two readings agreed and where they diverged.

  Day four: the records annex. Solt's committee file was not accessible through subsection eleven—that regulation covered formal correspondence, not committee membership records. But the committee's published mandate was public, and the mandate included a description of the informal review process. He read it twice. The language was careful, the kind of careful that had been written by someone who understood that the informal process needed to be defensible.

  *Member discretion in determining appropriate response.*

  Five words that did a great deal of work.

  Day five: Zone Three. He had the circuit mapped now—seven minutes at the corner pocket, four-minute recovery walking the corridor, seven minutes at the midpoint, four-minute recovery, five minutes at the far pocket. Total session: twenty-seven minutes of elevated saturation, twelve minutes of active recovery. He had run it three times this week. The efficiency drill numbers were improving at a rate he would not have achieved in the standard practice halls.

  He wrote that down without interpreting it.

  Day six: Aldea's office hours.

  Not to discuss the framework. Not yet. He had gone with a specific question about the calibration instrument's upper range—a technical question, narrow, the kind of thing you asked because you were working through the intermediate texts and had found a gap in the documentation.

  She had answered it.

  Then she had paused.

  Then she had said: "The research proposal with the faculty liaison. That's yours."

  "Co-signed," he said.

  Another pause.

  "The calibration study's original version is in the restricted archive," she said. Not to him—to a point slightly past him, in the way people spoke when they were confirming something to themselves. "The version in the intermediate section is the edited reprint. The reservations in the footnotes are condensed."

  He said nothing.

  "The original has six pages of supplementary data on the edge cases." She looked at him directly then. "The provisional access request—if it's approved—specifies which document?"

  "The original calibration study."

  She picked up her pen. Made a note he couldn't read from where he was sitting.

  "Your proposal should be approved," she said. "The documentation is complete." A pause. "I checked the faculty liaison's log this morning."

  He thanked her and left.

  He walked back to Block Four by the western route, not looking at the fourth floor of the academic wing, and wrote four lines in his notebook.

  Day seven: the fracture baseline.

  Fourteen days from the boundary crossing in the annex, he ran the same sequence: efficiency drill at seventy percent, six cycles, careful inventory at each release. The fracture was present. Unchanged—the same hairline quality, no deeper, no more prominent. Not progressing.

  He added a data point to the baseline log. One more week before he could make a determination.

  Day eight: combat theory class.

  Aldis was in the third row. Kael was in his usual position—fourth row, left wall, clear sightlines. Aldis did not look at him when he came in. Did not track him during the session. Operated exactly as he always did: attentive, precise, the model of someone who took everything seriously without appearing to try.

  Kael watched him for forty minutes between his own note-taking.

  The seminar with Solt was on Thursdays. This was Wednesday. Aldis had attended last Thursday—Kael had confirmed it from the academic wing's sign-in log, accessible through a different subsection of the records regulation, the one governing student access to facility use documentation.

  He had what he needed.

  Day nine: the library. The faculty liaison's log, visible through the standard request process, showed a status update on his and Sirin's proposal.

  *Under review. Estimated resolution: three to five days.*

  He had been expecting seven to ten.

  Three to five meant someone had moved it.

  He didn't know whether to file that as positive or negative yet. He needed more data.

  He went back to Zone Three and ran the circuit again.

  Day ten arrived.

  ?

  The formal assessment format had been posted four days earlier.

  Individual practical evaluation. Twelve students, single-round rotation, assessed sequentially. Scoring rubric: sustained output quality at declared load, response precision under interval prompts, and—new addition this rotation, not in the previous session's rubric—practitioner load at peak during the three-minute window.

  Peak load.

  Not average. Not declared.

  Actual peak, as measured by the platform inscriptions at maximum reading during the session.

  He read the rubric change three times.

  It was not targeted at him specifically—any change to the rubric applied to all twelve students in the rotation. But peak load was the metric that would reveal the difference between what he had declared and what he was capable of. In the last session, he had declared eighty percent and run at eighty percent. In this session, if he declared eighty percent and ran at eighty percent, the peak reading would confirm the declared value.

  But if he declared eighty percent and ran higher, the peak reading would show it.

  The question was what he was going to declare.

  He sat with that for two days.

  Then he made a decision, and on the morning of the tenth day he walked to the practice ground and waited for the session to begin.

  This novel's true home is a different platform. Support the author by finding it there.

  ?

  Aldis was in the rotation again.

  Position eleven. Kael was position six—the middle of the twelve, which meant he would assess before Aldis and Aldis would see his numbers before running his own.

  That was either an advantage or a disadvantage depending on what he declared.

  He watched the first five students run their assessments. The declared loads ranged from sixty-eight to eighty-two percent. Output quality was consistent with the declarations—no one was running significantly above or below what they'd put on the form. The peak readings matched the declared values within two or three points.

  The session was running clean. No anomalies. No one testing the boundaries of what the rubric was asking for.

  Position five was a student named Corr, Block Two placement, Stability 78. She declared seventy-seven percent and ran at seventy-nine. The assessor noted the two-point variance without comment and marked the form. Standard variation. Not flagged.

  Position six was called.

  He walked to the platform.

  He set his notebook on the bench.

  He had declared ninety percent.

  Not eighty. Not eighty-five.

  Ninety.

  The highest declared load in the session by a significant margin. The assessor had looked at his form twice when he submitted it and then looked at him once with the expression of someone recalibrating.

  He hadn't changed the form.

  He stepped onto the platform.

  The session started.

  ?

  Ninety percent felt different in a full practice hall than it did in Zone Three.

  Not the load itself—the load was the load, familiar, the vessel walls at full flex, the particular sharpening of attention that came with operating at the upper range of the first presence. He had run this level dozens of times in the annex and the maintenance corridor. He knew the texture of it as well as he knew anything.

  What was different was the room.

  Seven students still waiting their turn, watching from the bench seating along the wall. The assessor at his station, pen moving, recording the platform's inscription readings at intervals. The hall manager—a different one from the early-morning sessions, a woman who had been managing assessments for the Academy for longer than Kael had been alive—standing at the edge of the platform area with the practiced stillness of someone who had seen every variation of what students did under assessment pressure and had stopped finding any of it surprising.

  She was watching the platform inscriptions.

  He noticed that.

  He held ninety percent for the three-minute window without variation. Not the efficient-drill pattern, not the multiple-load-level test. Just sustained, clean ninety, the output quality as precise at the end of the window as at the beginning. The second presence was not engaged—it engaged at ninety-six, not here. He was running well within the first presence's range, which meant the output was clean in the specific way that came from a system that had no overflow to manage.

  At two minutes and forty seconds, the interval prompt came.

  Standard response drill: the assessor called a target load and the practitioner had to match it within two seconds.

  "Sixty-five."

  He dropped to sixty-five in a single clean transition. The platform inscription shifted color.

  The assessor noted something. "Back to declared."

  He returned to ninety. The inscription shifted back.

  The window ended.

  He stepped off the platform.

  ?

  Aldis was watching from the bench.

  Not obviously—the way Aldis watched things was the way careful people watched things, without their gaze drawing attention to itself. But Kael had been observing him for ten days, and he recognized the specific quality of Aldis's attention when it was directed at something that had adjusted his model.

  The assessment numbers were posted in real time on the board at the far end of the hall.

  Kael walked to the bench and sat down to wait for the remaining six positions.

  He had placed first in the rotation so far, by a margin that would hold unless someone in the remaining six declared higher and matched it.

  No one in the remaining six declared higher.

  ?

  Position eleven came.

  Aldis walked to the platform with the ease of someone who had done this many times, which was not accurate—he had done this twice, same as everyone else in the first-year cohort—but who had rehearsed the ease long enough that it had become a posture he didn't have to perform.

  He had declared eighty percent.

  He ran at eighty-one. Clean, precise, the transition prompt response exact to the tenth of a second. By any measure that the rubric used, it was an excellent assessment run.

  Peak reading: eighty-one.

  He placed second in the rotation.

  The margin between first and second was nine points of declared load, plus the output quality difference that came from running cleanly at a higher level rather than pushing against a lower one.

  He walked back to the bench.

  He sat.

  Kael was three seats down.

  Aldis looked at the board.

  Then he looked at Kael.

  Not the recalibration expression from the junction ten days ago.

  Something more deliberate.

  The expression of someone who had updated a model once already and was now updating it again, with more information, in a different direction than the first update had gone.

  He didn't speak.

  Neither did Kael.

  The session ended. The assessor collected forms. The hall began to empty.

  ?

  The western corridor outside the practice ground.

  Aldis caught up to him there—not by hurrying, by taking a path that converged on the exit Kael was using.

  Deliberate.

  "You declared ninety," Aldis said.

  "I did."

  "In the previous rotation you ran at eighty. The instruments confirmed eighty."

  "They did."

  Aldis looked at him. Not hostility. Not admiration. The specific expression of someone who was working through a problem that had more variables than they'd expected, and who was, despite the complexity, finding the process interesting.

  "You're going to run higher than ninety at some point," Aldis said.

  Not a question.

  Not an inference made tentatively.

  A statement of fact, delivered with the confidence of someone who had been watching carefully enough to know the shape of the thing even when they didn't have the numbers.

  Kael said nothing.

  Aldis stopped walking.

  They were at the junction of the western corridor and the path that led to Block One. Not the junction of the western route where they'd spoken ten days ago—a different junction, a different day, but the same geometry: two paths diverging, two people stopping at the point where the choice of direction became unavoidable.

  "Solt's seminar," Aldis said. "The theoretical work they cover." A pause—not hesitation, the pause of choosing precise words. "It assumes a ceiling. The dissipation assumption is foundational to the advanced curriculum."

  Kael remained still.

  "I've been in the seminar for four weeks," Aldis said. "I've read the materials. I've run the derivations." He looked at the board down the corridor, where the assessment numbers were still visible from here. "Ninety percent declared, run clean to the end of the window. No peak variance." Another pause. "You weren't at ceiling there either."

  The statement landed in the corridor between them and stayed there.

  "What do you want to know?" Kael said.

  Aldis looked at him directly.

  "Whether the ceiling is where they say it is."

  The question was precise. It wasn't asking for the number—Aldis already knew the number didn't tell him what he needed to know. He was asking something more fundamental: whether the framework was accurate, whether the map matched the territory, whether everything he had been building his understanding on for four weeks in Solt's seminar was a complete picture or a deliberate simplification.

  Kael looked at him for two seconds.

  "No," he said. "It isn't."

  Aldis nodded once.

  Not the acknowledgment gesture from ten days ago.

  Something that was, for the first time in their interactions, not calculated at all.

  He turned and walked toward Block One.

  Kael watched him for a moment.

  Then he turned toward Block Four.

  ?

  That evening, Room 208.

  The anomaly above the desk ran its minor irregularity into the quiet. He had stopped noticing it the way you stopped noticing sounds in a room you had lived in long enough—it was background now, part of the texture of the space, something that registered below the threshold of active attention.

  He opened his notebook.

  He wrote the assessment numbers first. Not because they required recording—the board had already done that—but because the act of writing them in his own notation put them in relationship with the other numbers he was tracking: the Zone Three saturation measurements, the fracture baseline, the calibration study's edge case documentation, the footnote he still hadn't traced to its source.

  The numbers told a story that the assessment board's record didn't have context for.

  He had placed first in the second formal rotation. He had done it at ninety percent declared load, which put his recorded ceiling in the assessment file at ninety. Not the instrument-error one hundred from enrollment, which was still flagged in his file as unresolved. Not eighty from the previous session.

  Ninety.

  The file would show a student who had performed at the top of his cohort at a declared load of ninety percent, with clean output and a precise interval response. The same student whose file had an instrument error notation and an administrative exclusion from the first rotation.

  The two pieces of information didn't resolve each other cleanly.

  That was intentional.

  He turned to the page where he'd been tracking the open threads.

  The fracture baseline: day seven. Stable. Seven more days to the determination point.

  The calibration study: proposal under review, three to five days, someone had moved it. He had, still, not determined whether that was Aldea acting through the faculty liaison or someone else acting for a different reason.

  The third thread—Aldis—was now different in character from how it had been when it opened at the junction ten days ago.

  He wrote three lines about it.

  Aldis had now, twice in ten days, stated an accurate inference about Kael's system directly rather than holding it. He had done this in the corridor tonight knowing that Kael was connected to a line of theoretical inquiry that Solt had publicly opposed nineteen years ago, and that Solt had his file on informal review. Either Aldis had calculated that this was a safe thing to do, or he had decided it didn't matter whether it was safe.

  Neither of those was the model Kael had been using.

  He underlined the three lines.

  Then he looked at the wall.

  Outside, the campus had entered its late-evening configuration. The rule field at three-quarters saturation, the academic wing's lights down to the single offices still occupied, the perimeter inscriptions running their maintenance pattern. Block One sat on the north end of the main quad, visible from this side of the campus if the angle was right.

  He was not looking at Block One.

  He was thinking about it.

  He closed the notebook.

  He lay down.

  He had moved all three threads forward in the last ten days. The fracture was stable. The study was in process. Aldis was something he didn't have a category for yet, which was the most interesting thing.

  He lay still and let it settle.

  Tomorrow: Zone Three at fifth bell, calibration study status check, combat theory class.

  And, somewhere in the day, a decision about what to do with the fact that the person who had most accurately read what he was, in the entire first-year cohort, was the same person who sat next to Solt every Thursday.

  He would not make that decision tonight.

  He didn't have enough data.

  He let sleep arrive.