The provisional access came through on a Tuesday.

  Three days, not five.

  He found the notification at the library desk before fifth bell—the clerk had placed it in the collection tray with a yellow tag that meant active processing complete. He had been checking the tray every morning for six days. On the seventh morning, the yellow tag was there.

  He collected the notification. Read it once, standing at the desk.

  *Provisional access granted: restricted archive, catalogue reference 7-A-19 through 7-A-22. Duration: forty-eight hours from date of issue. Access point: restricted reading room, second sub-floor. Accompaniment of faculty sponsor required on first access.*

  Faculty sponsor.

  He looked at the date of issue at the bottom of the form.

  Yesterday.

  The form had been processed yesterday and placed in the collection tray today. His forty-eight-hour window had started yesterday—which meant he had thirty-six hours remaining, not forty-eight.

  Someone had delayed filing the notification by one day.

  He put the notification in his notebook and walked to Aldea's office.

  ?

  She was already in when he arrived.

  The door was open—she left it open during the hour before first class, a practice he'd noted because it was inconsistent with how she operated during the day, when the door was closed and you knocked. The open door before class was an invitation of a specific kind: available, but occupied, which meant you could enter if your reason was sufficient.

  He knocked on the doorframe.

  She looked up.

  He held out the notification without speaking.

  She looked at it. Then at the date of issue. Then at the collection tray date.

  "When did you collect it?"

  "This morning."

  She was quiet for a moment. Not processing—she had already processed it. The quiet was the quality she used when she was deciding what to say, which was different from when she was deciding what to think.

  "I can accompany you before the first session," she said. "We have forty-five minutes."

  She stood, picked up the key that sat on the corner of her desk, and walked to the door.

  They went together.

  ?

  The restricted reading room was on the second sub-floor, below the main library stacks.

  Kael had been to the first sub-floor twice—the calibration archives he'd requested through Sirin's contact were housed there, and he had read the index without finding catalogue reference 7-A-19 through 7-A-22 anywhere in it. The index only covered the first sub-floor. The second sub-floor existed as a notation in the infrastructure records but wasn't accessible through the library's public-facing catalogue.

  He hadn't known what was down there until Sirin's research proposal had named the specific catalogue reference.

  He still didn't know how Sirin had known the reference number.

  He had filed that question under things to ask later.

  The reading room itself was smaller than he'd expected. Four tables, eight chairs, two standing shelves along one wall. A desk at the far end where a different clerk—older, with the particular attention of someone whose job was to watch what people did with the materials they were given—sat with a logbook open in front of him.

  The clerk logged Aldea's faculty credentials and Kael's provisional access number. He wrote both in the logbook with careful penmanship. He did not speak.

  He retrieved the documents from a locked cabinet behind the desk.

  Four folders. Catalogue references 7-A-19 through 7-A-22.

  He placed them on the table nearest the door and returned to his desk.

  Kael sat down and opened the first folder.

  ?

  The original calibration study ran to two hundred and sixty-three pages.

  The version he had read in the intermediate texts—the edited reprint—ran to forty-one.

  He had forty-five minutes.

  He did not try to read two hundred and sixty-three pages in forty-five minutes. He went to the supplementary data section that Aldea had mentioned—six pages, appended after the main analysis, flagged with the original researcher's handwritten note in the margin: *See appendix B for field-condition specifications.*

  He found appendix B.

  Eleven pages. Dense. The kind of technical documentation that had been written for an audience of one or two specialists and hadn't been edited for broader accessibility, because broader accessibility hadn't been the point.

  He read it in thirty minutes.

  The edge cases were real.

  Not theoretical. Not possible in principle. Real—documented, observed, recorded in detail by the original research team with notation of the specific field conditions under which each case had occurred, the practitioner involved in each case (identified by number, not name, which was standard research protocol for this period), and the precise measurement readings that had produced data inconsistent with the dissipation assumption.

  Case one: a practitioner whose Stability had read above the instrument's standard maximum in three consecutive tests under conditions of reduced ambient field pressure.

  Reduced ambient field pressure.

  He read that twice.

  Zone Three ran at reduced ambient field pressure relative to the practice halls.

  He did not note that yet. He kept reading.

  Case two: a practitioner whose output readings had continued increasing beyond the theoretical maximum output calculation, sustained for forty-two seconds before the researcher had stopped the test for safety reasons. The researcher's note: *Unable to determine cause. Instrument checked and confirmed accurate. Practitioner condition: normal. Repeating test with different instrument produced identical readings.*

  Case three: a practitioner whose field signature post-maximum-load showed a structural characteristic that the study's instruments couldn't classify under any existing framework category. The researcher had devoted three pages to trying to describe what they had seen, using analogies drawn from materials science and fluid dynamics. None of the analogies fully fit.

  Unauthorized content usage: if you discover this narrative on Amazon, report the violation.

  The note at the bottom of the case three documentation read:

  *We do not have the theoretical vocabulary to describe what we are observing. The dissipation assumption does not account for any of these cases. We recommend the assumption be reclassified as provisional.*

  The recommendation had been included in the full study.

  It had been removed in the edited reprint.

  He sat with that for thirty seconds.

  Then he turned to the last three folders—22 through 24—and found what he was looking for: the instrumental specifications that had been used to collect the data. The specific calibration parameters. The precise field conditions at the time of each observation.

  Case one had occurred under ambient pressure approximately equal to Zone Three.

  He wrote that down.

  Then Aldea said: "We need to go."

  He had seven minutes to first bell.

  He closed the folder.

  He had enough.

  ?

  Sirin was at the study tables that afternoon.

  She had seen the notification—he didn't need to tell her, because she had submitted the co-signed proposal and had her own access to the faculty liaison's status log. She had been waiting for him to come find her since the morning, which he could tell from the way she had arranged her texts. Three open, one closed on top. The same pattern as the first afternoon. But she wasn't working on a problem this time. She was waiting.

  He sat across from her.

  He did not lead with the documents.

  "The proposal access was issued yesterday," he said. "The notification was filed today."

  She looked at him.

  "The form is time-stamped," he said. "The issue date and the filing date are different by one day."

  "I know," she said. "I checked the liaison log this morning."

  He had expected her to have checked.

  "Who had access to the form between issue and filing?"

  "The faculty liaison's office processes them," she said. "The clerk, the supervisor, and any instructor who has the file flagged for notification." A pause. "Instructors on the Committee for Academic Standards can flag any file for notification."

  They both knew which instructor was relevant.

  "Solt was notified of our access before we were," Kael said.

  "Yes."

  He thought about that.

  Not angrily. The anger wasn't useful. What was useful was understanding the mechanism: Solt had seen the access request, known what it was for, been given one day's advance notice of their window. One day was not enough to act through formal channels—any formal response would have been visible in the record. But it was enough to act informally.

  To, for instance, alert someone who might benefit from knowing that a first-year student was about to read the original calibration study's full supplementary data.

  "Aldis," he said.

  Sirin looked at him. "You think Solt told Aldis?"

  "I think it's possible."

  A pause.

  "I think it changes the Aldis calculation," she said carefully. "If Solt told him as a warning—watch this student, here's what he's about to find—then Aldis's behavior in the corridor was more complex than it looked."

  "Or simpler," Kael said. "He might have decided the warning was information he could use in a different direction."

  Sirin considered this.

  She picked up her pen. Not to write—just holding it, the way she held it when she was working through something.

  "What did you find?" she said.

  He told her. Appendix B. The three edge cases. The reduced ambient pressure condition in case one. The researcher's recommendation that had been removed from the reprint.

  She listened without interrupting.

  When he finished she was quiet for longer than usual.

  "Case one," she said. "Reduced ambient pressure. You've been training in Zone Three."

  "I've been documenting it."

  "That's not documentation of a theoretical edge case." She looked at him. "That's a live observation under the same conditions the original study used."

  "Yes."

  "If you were to present that," she said, "with the Zone Three field measurements and the appendix B specifications alongside—"

  "It would require someone to explain why the recommendation to reclassify the dissipation assumption as provisional was removed from the published reprint."

  She set the pen down.

  "That's not a question the examination board wants asked," she said.

  "I know."

  "It's also not a question that can be suppressed indefinitely if the documentation is complete." She looked at the table between them for a moment. "How complete is it?"

  "Getting there." He paused. "The fracture baseline needs seven more days."

  She looked up. "The fracture is connected to this?"

  "I don't know yet. Case three in the appendix—the unclassifiable structural characteristic. I may have something analogous." He said it the same way he said everything—not performance, just accuracy. "If the fracture is what I think it is, it's evidence of a practitioner operating beyond the dissipation assumption's boundaries and sustaining a specific class of load event that shouldn't be possible under the current framework."

  Sirin was very still.

  "You went past the second presence's ceiling," she said.

  He had not told her about the second presence explicitly. He had told her about the annex. He had told her about the training hall. The inference was available from what he'd shared, and she had made it.

  He filed that.

  "Once," he said. "Documented. Not repeating until the baseline is clear."

  She nodded slowly.

  Not the quick acknowledgment of someone moving past information to the next point. The slow nod of someone integrating something that changed the weight of things they'd already been holding.

  "Seven more days," she said.

  "Seven more days."

  She picked up her pen again. Opened her notebook to a fresh page.

  "Then we write the research application," she said. "Not the provisional access request—that was stage one. A formal research application to the examination board, through Aldea, for access to the full instrumental specifications and the practitioner records from the original study." She began writing. "The application requires a preliminary finding. What we have is already sufficient for that. What you have in Zone Three plus appendix B's specifications is enough to establish that the edge cases are reproducible."

  "The examination board will send it to the committee," Kael said.

  "Yes. And the committee includes Solt." She kept writing. "Which means we need the application to be so complete that the committee can't decline it on documentation grounds. It needs to be declined on substance or not at all." She looked up. "Solt declined Aldea on substance nineteen years ago. He said the theoretical case was interesting but served no productive institutional function."

  "He can't say that with a live reproducible case."

  "He can try." She held his gaze. "But trying will require him to engage with the evidence. And engaging with the evidence means the evidence becomes part of the formal record."

  The formal record was what the informal process had been designed to avoid.

  He understood what she was building.

  Not a confrontation. A mechanism.

  A process so formally complete that the only available responses were either to engage with it on its own terms—in which case the evidence became visible—or to decline it in a way that was itself documentable.

  "You've been thinking about this since the proposal," he said.

  "I've been thinking about it since the corridor," she said. "When you told Aldis the ceiling isn't where they say it is."

  He looked at her.

  "I was at the junction across the courtyard," she said. Not explaining herself—just providing data. "I heard it."

  He sat with that.

  She had been in the courtyard. She had heard the exchange. She had spent two days building the next stage of the strategy before he'd arrived at the study tables.

  He had been running the Aldis thread and the calibration thread as parallel problems. She had been running them as a single problem.

  "Seven days," he said.

  "Seven days." She turned to the next page in the notebook. "In the meantime—the Zone Three measurements. I need all of them. Dates, saturation readings, load levels, session durations. Everything."

  "I'll get you the notebook."

  She looked at him once more.

  Not revising an estimate this time. Not taking a measurement.

  Just the look of someone who was in the same room as a problem they intended to solve, and who had recently confirmed that they were not solving it alone.

  She turned back to the page and kept writing.

  Outside, the afternoon light held steady. The campus continued around them in its usual configuration—scheduled, indifferent, running on the logic it had run on for seventeen years.

  They worked in the silence of people who had sufficient material and were building something with it.

  ?

  He took the western route back at the eighth bell.

  The campus in early evening ran at the specific temperature of the season's last warmth—stone releasing the heat it had accumulated over a day of sun, the air cooler than the ground, the rule field at its pre-nighttime saturation level, still elevated from the day's activity.

  He thought about the three things he'd confirmed today.

  The original study's supplementary data was what appendix B had described it as. The edge cases were real and documented. The conditions of case one matched Zone Three within acceptable measurement tolerance.

  Solt had been notified of their access before they were.

  Sirin had been in the courtyard.

  The first was a structural fact about how the system worked: informal access to formal processes, the same architecture he had been reading since the records annex and the committee correspondence. It was not a new fact. It was the same fact, confirmed again.

  The second and third were new data.

  He revised the Sirin model again. Not the capability estimate—that had been accurate from the beginning. The scope estimate. He had modeled her as working in parallel on adjacent problems. She had been working on the same problem. The distinction was larger than he had been treating it.

  He stopped at the entrance to Block Four.

  The building was lit in its early-evening pattern, the sounds of students moving between rooms, someone on the upper floor playing a string instrument quietly, badly, the specific quality of a person who was learning rather than performing.

  He stood there for a moment.

  Seven days to the fracture determination.

  Seven days to the research application.

  He went inside.