The Hall Registers

I had left the observer-confirmation section open in my index overnight, still unresolved, and the cohort session pulled me out of the hall the next morning before I had decided what to do about it. The session ran from second bell to just past fifth, and by the time I got back to the unlabeled hall my notes were already three pages behind what had actually happened. I sat down at the table, smoothed the top sheet, and tried to write while the details were still precise.

The exercise had gone the way I expected, which was both good and slightly unsettling. I had gone with item fourteen, Temporal affinity interaction, because the structural account it required mapped cleanly onto what I had already drafted, and because the detectability conditions were specific enough to be useful to a board reviewer without being so narrow that they looked like fishing for a particular conclusion. The restricted candidate version's additional requirement, the part where I had to indicate my own configuration's relevance, I wrote the way I would have written any documentation: accurately, without editorializing, citing the two system-logged properties by their logged designation rather than by my own phrasing. What happened in the board file from that point was not mine to manage.

My lamp still sat at three-quarters reservoir from last night, so I refilled it anyway, tipping the oil tin until the level climbed to the upper mark, then replaced the cap and put the tin back in the cupboard below the shelf. The habit had started after day three, when I had been three pages into a footnote chain and the light guttered at the worst possible angle. Now I checked it every time I sat down, the way you check the window latch in a city apartment even when you know you closed it.

Nothing had changed on the INTERIM TRAINING ACCESS (LIMITED) panel, the same line it had shown since partial attunement status engaged: theory and system literacy modules only, no affinity-framework-dependent content. Twelve days and a handful of hours remained in the Verath Regional window. Two system-logged properties sat unconfirmed, both waiting on a Council-level observer who had not appeared in the hall for four days, someone the module documentation did not explain, either in terms of timing or in terms of what happened to the confirmation pathway if she never came.

I opened the restricted theory index and queried COMPOUND-PRIMARY TRAINING ENVIRONMENTS. The index responded in its usual form: a text prompt rather than a direct result, asking me to specify what I already knew about the subsection topic before providing access. That particular exchange had gotten old around day six, when I realized the index would not accept uncertainty as an answer. Writing "I don't know what this covers" looped it back to the same prompt. Writing a summary of adjacent material I had already covered made it move. So I wrote that compound-primary candidates operated without a labeled training context and that the physical architecture of an unlabeled hall might differ functionally from a standard hall, and after a brief hold the index served up a subsection I had not seen before: COMPOUND-PRIMARY TRAINING ENVIRONMENTS: PHYSICAL ARCHITECTURE.

The subsection had loaded without a query prompt at the end, which was unusual. The index had produced it as a response to what I had written, but the timing was different from its usual pattern, where a second confirmation step always occurred before new material loaded. This had simply appeared. I read the first paragraph twice before continuing.

Design principles for unlabeled training spaces filled the header section. Standard halls, the subsection said, calibrated to a candidate's confirmed affinity type and configured with materials and objects that constituted exercise prompts, items the candidate could interact with directly to practice and extend their affinity's active expression. An unlabeled hall, by contrast, calibrated to nothing confirmed, because there was nothing confirmed to calibrate to. Instead, it calibrated to the candidate's current registration state.

That was the word it used: registration state, not attunement state, not affinity expression, not classification. In the context of the documentation I had been reading for twelve days, it was the specific term for the composite status a candidate carried in the system's active records at any given moment: the sum of their partial attunement data, their logged properties, their unresolved classification flags, their outstanding review statuses. My registration state was, depending on how you counted, one partial attunement, one UNRESOLVED (ACTIVE) affinity flag, two system-logged inference properties both unconfirmed, and one ongoing board review with twelve days remaining.

The next paragraph described how an unlabeled hall's physical objects functioned. In a standard hall, a pedestal object was a prompt: touch it, use your affinity on it, respond to it. In a compound-primary training space, pedestal objects were registers. They did not wait for the candidate to act on them. They reflected the candidate's registration state back in physical form, translating abstract classification status into observable configuration. I put down my pen and looked at the objects on the pedestals.

Four objects occupied the pedestals. I had catalogued them on day one out of thoroughness and had not touched any of them since, because nothing in the LIMITED access modules had authorized engagement with physical exercise materials. The leftmost pedestal held a smooth stone tile. Next to it sat a glass cylinder sealed at both ends, containing a dark liquid that never fully settled even when tilted. The third pedestal carried a metal disc balanced on a thin vertical pin, with its rim touching nothing and its weight resting only on the pin at the exact center point. On the last one, a short length of cord looped through two anchor rings that kept it from going slack or taut, its tension resting somewhere between the two.

The disc had remained balanced since day one, and I had interpreted this as the object's default state, the way you assume a clock on the wall shows the right time until it proves otherwise. The subsection said the objects reflected suspended classification states, and the subsection mentioned the disc specifically, not by name but by type: a balanced point register, suited to representing configurations that exist at a boundary between classification outcomes. My configuration sat at a boundary, and the partial attunement process had produced that exact administrative condition rather than a way of describing it. I had been sitting three feet from a live registration feedback instrument for twelve days without knowing it was one.

Movement caught at the edge of my vision, and the disc had shifted. It was not a large movement, nothing that would have caught my attention if I had not been looking directly at it. The rim had tipped maybe four degrees clockwise. The pin still held it at center, but the point of balance had visibly changed. No draft moved through the hall. The nearest window was closed and latched. Nothing had shifted nearby that I had touched. The index had no active exercise prompt running, and nothing I could identify as a cause was present.

I wrote: disc moved. Four degrees clockwise. No contact, no wind, no active prompt. 9:40 post-session, second day of fourteen-day window. Then I sat still and thought about what the subsection had just told me.

If the objects registered the candidate's state rather than responding to active expression, a change in state would produce a change in the object. My state had not changed in any obvious way. No new system notifications, the status panel looked exactly as it had at the session this morning, and the fourteen-day window was counting down at the rate it always counted. Except that the conceptual mapping exercise might already have registered somewhere. The restricted candidate version produced a candidate-authored document that entered the board file through the may-be-cited mechanism. If that document had gone in, the system had processed a new entry into my registration record: not a classification change, not a resolution, just an update. But an update was still a change in registration state.

I opened the status panel to the PARTIAL ATTUNEMENT STATUS entry and read through the text carefully. It said what it had always said: attunement process produced partial data, classification review ongoing, see Verath Regional tracking number. The disc did not move. I closed that entry and navigated to AFFINITY: UNRESOLVED (ACTIVE).

The disc moved the same direction as before, another three or four degrees clockwise, slow enough that I could track it without blinking. Then it held. I closed the status panel entirely, and the disc held its new position.

I wrote: PARTIAL ATTUNEMENT STATUS: no disc response. AFFINITY: UNRESOLVED (ACTIVE): disc moved, same direction, approximately equal magnitude. Closed panel: disc held. Not returning to prior position.

The implication was mechanical once I laid it out. The disc did not track what I was reading in the abstract; it tracked which specific status flag was active in my current focus. PARTIAL ATTUNEMENT STATUS was a container, a label for the situation without the live flag underneath it. AFFINITY: UNRESOLVED (ACTIVE) was the flag itself, the thing the system still processed. The disc responded to the live flag. That was something the subsection had not spelled out directly. It had described the objects as reflecting registration state, but had not specified that the response was granular down to the individual flag level. I had inferred that from two data points, which was the minimum necessary for inference and also the minimum that frequently proved wrong.

I needed to test the inference properties separately. Status panel open, I navigated to the compound-primary inference log, where the two system-logged properties sat: COMPOUND-PRIMARY RESONANCE PROPAGATION: INTERACTION WITH TEMPORAL AFFINITY STRUCTURES NOTED, and CROSS-DETECTION PATHWAY VIA ANALYTICAL PRE-EVENT WINDOW NOTED. Both read UNCONFIRMED, both still waiting on the three-condition observer protocol. The disc did not move. I held the panel open on those two lines for a full minute, watching. Nothing happened.

I wrote: inference property lines: no disc response. Null result. Hypothesis: disc responds to classification-state flags (UNRESOLVED, ACTIVE) but not to logged-inference properties, which may occupy a different status category. Logging as data point, not failure.

The distinction mattered because it suggested the disc was tracking something more specific than "everything in my file." The UNRESOLVED (ACTIVE) flag was a classification state, an active system process. The logged inference properties were candidate-submitted theoretical materials, held pending external confirmation. Same file, different category, and the disc apparently cared about one and not the other. This was new information, not about my attunement configuration or the confirmation protocol, but about the hall itself and how it registered the difference between things the system was actively processing versus things it had filed for later.

Someone knocked at the doorframe.

Tobias was standing at the entrance to his training hall, which shared a wall with mine, his door open behind him. He wore the same plain session clothes he always did, hair still damp from something, and he had the look of someone who had finished a good portion of his afternoon work and now allowed himself a pause.

"Question," he said.

"Go ahead."

"The theory track." He leaned against the doorframe. "Does it have a ceiling?"

"Yes," I said. "I have hit it."

He did not answer right away, though it did not read as hesitation. He was already working out his next question.

"What comes next?"

"Observer confirmation. Then Verath Regional review."

The pause after that was about two seconds. I counted, not because it mattered, but because timing was one of the things Tobias communicated in when he chose not to use words.

"I did not know the system built external dependencies into individual candidate tracks," he said.

"Apparently it does."

He shifted his weight, a small adjustment that looked like the extra beat before a decision. "If the observer does not come within the window, what happens?"

"That's in the documentation," I said. "The board extends the provisional classification status until a review can be scheduled with a different observer. The training access lapses during the gap."

He nodded, not in agreement exactly, more acknowledgment of the information and its implications. Then he went back through his own door. The latch caught behind him with a solid mechanical click the door produced at about thirty degrees from fully closed.

I sat looking at the space where he had been standing. Tobias had asked the question casually enough that a less careful listener would have logged it as small talk. He had wanted to know whether my progression had a structural limit, and the follow-up asked what that limit looked like. Now he knew it required external participation from a party outside my control, which was information about my relative position in the cohort that he had not had this morning.

Five candidates worked in labeled halls, making documented progress at whatever rate their affinities and hall configurations allowed. Renne's Force track alone had logged forty-seven boundary flags that session. Nadia stayed deep in her Analytical work. Petra and I remained in deferred or restricted status. The cohort session had happened, and the fully attuned candidates had moved through it differently than the restricted candidates had, because the mapping exercise replaced the calibration component for us, and the exercise outcome entered different administrative channels. Tobias had asked about the ceiling before the session results would have had time to propagate to his panel. He had been curious before confirmation.

I wrote that down too, in the section of my notes I kept for non-system observations, then looked at the four pedestal objects and the disc in its new position, seven or eight degrees clockwise from where it had started twelve days ago. The other three objects had not moved while I had been running the status panel tests. The stone tile sat flat and unchanged. The cord held its intermediate tension without any visible shift, and the glass cylinder with its dark liquid looked the same as always, neither fully settled nor agitated, fixed in that in-between state the way it had been since day one.

The subsection had not described what each object type registered specifically. It had described the category, classification state, and left the specifics to contextual inference. The stone tile might track something entirely different from the disc, might respond to a different flag or a different category of information in my file. Or it might not respond to my current state at all, might be waiting on a state change that had not happened yet.

I queried the index again, this time asking specifically about the stone tile and cord configurations as compound-primary registers. The index asked what I already knew about them. I wrote out a summary: one confirmed data point for the disc responding to AFFINITY: UNRESOLVED (ACTIVE), a null result for the disc and the inference properties, and no observations yet for the tile or cord. The subsection described the objects as classification state registers without specifying which object tracked which category.

The index returned a short prompt: PHYSICAL REGISTER CONFIGURATIONS VARY BY PRIMARY AFFINITY TYPE AND COMPOUND CLASSIFICATION STATE. SPECIFIC CONFIGURATIONS DEPEND ON HALL INITIALIZATION AND THE THEORY INDEX DOES NOT PROVIDE THEM. HALL-GENERATED OBSERVATIONS CONSTITUTE THE PRIMARY INFORMATION SOURCE FOR THIS CATEGORY.

So the index would not tell me. The hall itself was the documentation. That was consistent with everything else about this training space. The index worked through text exchange and confirmation steps. The hall worked through direct physical response. They were parallel channels, and the hall's channel was the one that did not require me to demonstrate prior knowledge before giving me something.

I put my pen down and looked at the other three objects. Outside the window, the afternoon light had shifted enough that the far wall of the courtyard had gone from full sun to partial shadow, which meant it was getting close to the fourth bell. The session had run most of the morning, and I had spent another hour writing notes before sitting down with the index. The day was mostly gone.

The tile was smooth-surfaced grey stone, about the size of my palm, sitting flat on its pedestal without any obvious orientation marker. The cord was a braided fiber, light brown, looped through two small iron rings mounted to a short wooden frame, its tension sitting at roughly midpoint. Both had sat on their pedestals since day one, motionless throughout every session I had spent in this hall.

The disc was no longer motionless. It sat seven or eight degrees off its original position, holding there without any apparent support structure change, because the pin at its center was the same pin, and the disc's weight had not changed, and the only variable was whatever the hall tracked in my registration state.

I wrote: stone tile: no observations. Cord: no observations. Will test with status panel navigation as I did with the disc. Methodology established. Then I noted the session log was still open in my panel, as it had been since the cohort session ended, and that the disc's movements were now part of the session record whether or not any administrative pathway connected them to a formal exercise category. I had not opened the session log intending to document the hall behavior. The session log had been open because I had been writing notes from the morning exercise, and the two had overlapped. If a Council-level observer reviewed the session log, that record would be there. If Verath Regional pulled the session data as part of the classification review, it would be there. I had not staged any of this. The hall had moved, I had observed, I had written it down, and the log had caught it.

The stone tile sat on its pedestal. The cord held its middle tension between taut and slack. The disc was at eight degrees clockwise.

I opened the status panel, navigated to AFFINITY: UNRESOLVED (ACTIVE), and watched. The disc held steady, already reading the flag. I closed that entry and navigated to PARTIAL ATTUNEMENT STATUS, then held it there, watching the tile. Nothing happened. I closed that and opened the inference log, watching the cord, and got nothing there either.

The cord and tile either corresponded to status categories I had not isolated yet or they were waiting on state changes that had not occurred. Both were possible. The cord's braided linen was wound tight around the anchor rings, the kind of weave that held its shape even under steady tension. I tried a different entry, the ADMIN section that listed the EXTENDED SESSION AUTHORIZED flag, navigating there from the main menu. The cord did not respond. I closed the panel.

I had established the methodology, as I had written: test each object against each navigable status entry, record responses and null results with equal attention, treat the hall as the documentation it apparently was. Twelve days remained in the window. No Council-level observer had appeared. Two unconfirmed properties sat in the log, stable and waiting. One disc held a new position, eight degrees from where it started, sitting there without anything external to explain it, tracking something internal that the hall could read and I was still mapping.

The session log was open. The lamp was full. I picked up my pen and started on the cord.