The second combat simulation rotation ran on the twelfth day.

  Kael arrived forty minutes early.

  Not because he was anxious.

  Because the practice ground had a specific acoustic and field distribution profile he hadn't trained in yet. Forty minutes was enough time to walk the assessment platforms, feel the saturation gradient at each one, and identify which position gave the cleanest reading.

  He wasn't going to ask for a specific platform.

  He was going to let the assignment process run normally and know in advance what he'd been given.

  The platform assignments were posted outside the hall manager's door at thirty minutes before the session.

  He checked the list.

  Platform seven.

  He had already identified platform seven as the second-best position in the hall—slightly lower ambient saturation than platform four, which was where the two main lattice runs crossed, but more consistent field distribution across the surface. Platform four had a micro-variance at its eastern edge that would have introduced a variable into his output readings he didn't need.

  Platform seven was cleaner.

  He went inside and waited.

  ?

  There were twelve students in the second rotation.

  He had looked at the list the morning it was posted and identified each name against what he'd observed over the past twelve days.

  One name he'd noted specifically: Aldis.

  Block One placement. Second in the first rotation—high enough to be significant, not so high as to attract the specific attention that came with ranking first. He moved through the corridors with the ease of someone who had already decided where he stood and found the decision confirmed.

  He was also, Kael had established through two weeks of environmental observation, the person who sat next to Instructor Solt at the advanced seminar on the fourth floor of the academic wing.

  The seminar was closed to first-year students.

  The regulation was clear on this. Advanced seminar access required second-year standing, performance metrics in the top quartile of the prior year's cohort, and instructor nomination.

  Aldis was a first-year student.

  He had noted that Aldis attended it.

  He had noted, additionally, that Aldis had not applied for an exemption through the standard process—because the standard process was documented, and the records annex would have shown it, and he had checked.

  Which meant the access was informal.

  Informal access, granted by an instructor who sat on the Committee for Academic Standards and Integrity.

  He had drawn no formal conclusions.

  He had a data point.

  ?

  The session format was straightforward: six pairs, sequential assessment, each student evaluated on sustained output, Stability under load, and response speed across a standard three-minute interval.

  The pairs were randomized.

  The rotation supervisor drew names from a physical box, which Kael had been watching since he arrived to confirm there was no mechanism for non-random selection built into the process.

  There wasn't.

  His pairing was random.

  He was assigned to run third.

  He watched the first two pairings from the waiting area.

  The first pair performed at the level he'd expected—mid-tier, both of them, competent and unremarkable.

  The second pair was more interesting: a student named Wren, whom he'd assessed as high-tier based on her efficiency of motion in the corridors, against a student named Dalt who had the particular tension in his shoulders that came from someone who had been training hard for a specific performance. Wren performed cleanly. Dalt overextended on his third interval and his Stability reading dropped four points in the final thirty seconds.

  The assessor noted it.

  Kael noted that the assessor noted it.

  His name was called.

  ?

  His pairing was against a student named Coriss.

  Block Two. Upper-mid tier—not the highest, but selected for. She had ranked fourth in the first rotation. Focused, unhurried quality that Kael had observed twice in the library and once on the western path—the signature of someone who had developed good technique over a longer timeline than most students her age.

  She was not going to be easy.

  He had not expected easy.

  They took their platforms. The assessor confirmed the parameters. The session began.

  ?

  He loaded to eighty percent.

  Not one hundred. Not ninety-six, where the second presence engaged. Not even the eighty-five he'd been running in Zone Three.

  Eighty percent—above the standard training load for a first-year student by a significant margin, clearly competitive, entirely within the range he could sustain for a three-minute interval without any involvement of the second presence, without any fracture-related complications, without anything the assessor's instruments would not have a clean and simple explanation for.

  He had spent two days deciding on eighty percent.

  The decision was not about hiding what he could do.

  It was about controlling what he revealed, and when, and to whom.

  The instrument error notation was already on his file. Solt was already monitoring his enrollment status. The first combat simulation had produced an administrative exclusion from what was meant to be his baseline reading. The research proposal with Sirin was in process but not yet resolved.

  If you come across this story on Amazon, be aware that it has been stolen from Royal Road. Please report it.

  He had three open threads with uncertain outcomes and a fourth thread—the fracture—that he was still characterizing.

  Showing one hundred percent in the second rotation, in a session that contained Aldis, in a hall whose records fed directly to the examination board—that was not the correct move. It was the kind of move that felt decisive but produced consequences faster than he could manage them.

  He knew what correct looked like.

  It wasn't making an argument he wasn't ready to finish in a room he hadn't chosen.

  Eighty percent was a controlled variable.

  It demonstrated what it was designed to demonstrate and nothing else.

  ?

  He held the load for the full three-minute interval.

  His Stability reading ran at ninety-four percent across the interval—no error codes, no ceiling flags, nothing unusual. His output rating placed him in the top tier of the session. Coriss's Stability reading had dropped to eighty-seven in the final minute. Still respectable. The differential was clear.

  He released cleanly on the assessor's signal.

  Stepped off the platform.

  Went back to the waiting area and watched the remaining pairs.

  Aldis ran fifth.

  Precisely. Cleanly. The technical control of someone whose technique was automatic. His Stability reading held at ninety-two percent across the full interval. His output matched his placement tier.

  He also glanced at Kael once during the interval.

  Not at the end—during it, at approximately ninety seconds in. In the way of someone running a performance they'd prepared and wanting to confirm a specific data point was still in the room.

  Kael was still in the room.

  He noted the glance and filed it.

  ?

  The rotation results were posted that afternoon.

  He ranked first. Aldis ranked second.

  The gap between their output scores was not large—three points, within normal variation. But it was visible and it was in the record.

  He stood at the posting board for ten seconds, confirmed the numbers, and went to the annex.

  ?

  The annex was dark at fourth bell, the light through the high window flattening into the diffuse gray of late day.

  He sat on the floor and did nothing for five minutes.

  Not thinking—he was past the point where he needed to sit with his results to understand them. He was simply letting the assessment run out of his system the way effort ran out.

  Cleanly. Without ceremony.

  Then he opened his notebook.

  Half-load check before curfew. Sixty percent, four cycles, no differential points, just baseline monitoring. He had not run his system at assessed levels in a practice ground since the training hall incident weeks ago, and he wanted to verify that performing in a formal environment—with instruments tracking output in real time—hadn't altered the fracture's behavior in ways he hadn't observed in the annex.

  He loaded to sixty percent.

  Clean. The fracture present at its usual perceptibility level—the hairline in the background, the settled quality he'd learned to read as stable.

  Nothing different from yesterday.

  He held for two minutes. Released. Reloaded.

  On the fourth cycle, he felt something.

  ?

  Not the fracture changing.

  The second presence.

  Not engaging—he was nowhere near the threshold. But there was a quality at the edge of his Capacity's upper range that hadn't been there before the practice ground assessment. Not a problem. Not instability.

  A kind of heightened readiness.

  As if the performance at eighty percent had brought the upper range of his system into sharper awareness of itself.

  He held the load and paid attention.

  After forty seconds he had a working description: the second presence was more perceptible at lower loads than it had been two weeks ago. Still inert below ninety-six percent. But more present as a background element.

  The way a door in a familiar room became more noticeable once you'd opened it and knew what was behind it.

  Not the fracture—the fracture was a change in the layer below function. This was a change in the layer above it. In the texture of his awareness of his own system, rather than in the system itself.

  He wrote for eight minutes.

  *Second presence: perceptibility increase following formal assessment. Not a function of load—possibly a function of extended sustained performance in an observed context. Hypothesis: the formal assessment environment, with external instruments tracking output in real time, created a feedback condition that heightened internal awareness of the system. Not mechanical. Not measurable externally. Present and consistent across four reload cycles.*

  A pause.

  *Alternative hypothesis: cumulative effect of Zone Three training over twelve days. The differential saturation work may be sensitizing the system's self-monitoring function. Cannot separate variables yet. Run longer baseline.*

  He put the pen down.

  He had mapped Zone Three and established a training regime he could sustain without visibility. He had identified Solt as the origin point of the administrative intervention. He had co-signed a research proposal with Sirin that was currently in process with the faculty liaison.

  He had also, this afternoon, placed first in the second combat simulation rotation using a controlled load that was not his maximum.

  The record would show what it showed.

  Anyone reading it would see a Block Four student who had performed at the top of his assessment group with a Stability reading of ninety-four percent, clear output metrics, no error flags.

  They would not see that he had held the performance below what he was capable of.

  They would not see the second presence, inert at eighty percent, waiting.

  They would not see the fracture, still present and still stable, running its slow fourteen-day baseline in the background of everything he did.

  They would see the number.

  That gap—between what the number showed and what was actually present—was, for now, entirely his.

  He put the notebook away.

  ?

  He was on his way back from the annex when he encountered Aldis.

  Not by plan—the western path was his standard route back to Block Four, and Aldis was coming from the opposite direction. They arrived at the junction near the main courtyard at approximately the same moment. The geometry of the path meant that passing without acknowledgment would require one of them to alter course.

  Neither of them altered course.

  They stopped at roughly two meters' distance and looked at each other.

  The late afternoon light came from behind Kael, which meant he could read Aldis's expression and Aldis was reading him against the light. He noted that. He noted also that Aldis hadn't altered course—which meant either he hadn't planned the encounter, or he had planned it and was performing not having planned it, and both of those were possible.

  He would not determine which from this interaction alone.

  He noted both possibilities carefully and continued walking.

  Aldis was taller than him by four or five centimeters. His Block One placement was visible in the cut of his robe—standard Academy issue, but the Block One allocation ran slightly better materials. The difference was perceptible if you were looking for it.

  "Block Four," Aldis said.

  Not a question. A confirmation of a fact that was also, in the person saying it, a statement about the significance of the fact.

  "Block One," Kael said.

  Aldis looked at him for a moment.

  Not hostile.

  More like the expression of someone who had been given a result they hadn't quite prepared for and were deciding how to process it. The combat simulation scores had been posted for three hours. He had had time to see them.

  "You ran below your ceiling," he said.

  Not a question either.

  An inference—clean, accurate, delivered without flourish.

  Kael registered that it was accurate and that Aldis had made it quickly, from a three-minute performance interval, in a shared session where they had not been paired against each other.

  Observant.

  He updated his model.

  "The instruments read what they read," Kael said.

  "The instruments read ninety-four percent Stability at eighty percent load." Precise. He'd looked at the detailed output numbers, not just the ranking. "You weren't at ceiling."

  Kael said nothing.

  Aldis looked at him for another moment.

  Then something shifted in his expression—not warmth, not hostility.

  The specific recalibration of someone who had been operating from one model and had just encountered data that required a different one.

  "The next formal assessment is in ten days," Aldis said.

  "I know."

  Aldis nodded once—a gesture that was less agreement than acknowledgment—and walked past him.

  Kael stood at the junction for two seconds.

  He had updated the model twice in the last ninety seconds.

  First update: Aldis was more observant than his second-place ranking had suggested. To read a performance ceiling from a three-minute interval in a session where they hadn't been paired required both the technical knowledge to understand load levels and the attention to apply that knowledge in real time to someone else's output. Most students couldn't do that. Most students were focused on their own assessments.

  Aldis had been running his own assessment of Kael during his own assessment.

  Second update: Aldis had stated the inference directly rather than holding it. That was a choice. He could have held it—filed it, let it accumulate value as private information. Instead he had stated it at the junction, in the moment of passing, which suggested either that he wanted Kael to know he'd read it, or that he wanted to see how Kael responded to being read.

  Both were interesting.

  He filed both.

  Then he continued toward Block Four by the western route.

  Behind him, the academic wing's lights were coming on in the early dark, floor by floor—systematic, indifferent, running on timers that didn't account for whether anyone was still inside.

  On the fourth floor, Instructor Solt's office window was lit.

  He did not stop to look at it.

  He already knew what he needed to know.

  The rest would arrive at its own pace.

  The notebook in his pocket had two open threads, and a third had just opened at the junction of the western path—smaller than the others, less certain, but present.

  He had twelve days before the formal assessment.

  He intended to use them.