Marcus woke at 5:47 AM on September 17, without an alarm.
This was not unusual. He had been waking at this hour since August, without planning to. The apartment was quiet at 5:47. Outside, the city was beginning its early-morning processing: delivery cycles, industrial ventilation systems, the transport network's first-hour reconfiguration. He had learned to distinguish these from each other by sound over three and a half years of living in proximity to the Branch 12 monitoring cluster. The cluster did not make sound. He had listened for it anyway, in the beginning.
He made coffee and stood at the window.
Forty-three months ago, when he had taken the dormant-baseline monitoring assignment, he had understood it as a decision to watch without accumulation. The Traced system offered more active monitoring roles — contextual weight assessment, interpretive flagging, pattern attribution — all of which required the observer to engage with the system's interpretive layer. He had chosen dormant-baseline precisely because it did not. You checked the number. The number meant what it meant according to the system's own parameters. You did not add to it.
This had seemed simple at the time. He had maintained it. He was not certain it had stayed simple.
The novel was on the table. He had finished it at 10:15 the previous night and put it down without marking the page. There was no page to mark; it was the last one. He had read the ending twice: the appointment-book man, at the close of the final chapter, had five appointment books open simultaneously on the desk in front of him, each one opened to a different month, and he had not closed any of them before the book ended. The consolidation the reader had been expecting since chapter three had not happened. The character sat with five books open and the novel ended. Marcus had put the book down and gone to sleep.
He was thinking about the five books now, at the window, with his coffee.
The meeting was at 10 AM. Item 4b of the agenda was titled, in the documentation he had received two weeks ago: Proposal for Suspension of Dormant-Baseline Monitoring Protocol, Branch 12 Subset. He had read the proposal twice when it arrived and filed it. He had not annotated it. He had been annotating the Branch 12 logs for forty-three months, but the proposal was not a log. It was a recommendation. He did not annotate recommendations.
Branch 12 was currently at 87 percent. He had checked on Tuesday, which was the designated check date, and the trace index had been 87 percent, up from 84 percent the week before. Three points of upward drift in five days. Still within normal variance range. Still consistent with the dormant-baseline pattern he had been observing since February of three years prior. He had noted it without interpretation. The Traced system's contextual weight assessment feature had flagged the reading as within expected parameters and offered a contextual note. He had read the note and closed the dashboard.
He had not opened the dashboard this morning. The meeting was four hours away and the 87 percent was the number he was bringing.
The Branch 12 monitoring cluster tracked what the Traced system documentation called trace density in the secondary attribution layer — a category of cognitive weighting that the system's original architects had identified as statistically significant in predicting behavioral pattern shifts, but that the interpretive advisory group had argued was too context-sensitive for automated flagging without human baseline assessment. The dormant-baseline protocol was the resolution. A human observer would watch the number without the system telling them what the number meant.
The number had meant different things to Marcus at different points in forty-three months. In year one it had meant data. In year two it had meant pattern. In year three it had started to mean something he did not have a clean word for: a posture, maybe. The sustained attention of someone who had been told not to interpret but who had been present long enough that the absence of interpretation was itself a stance. He had thought, in year three, that he should probably report this to the interpretive advisory group. He had decided this was outside the scope of the monitoring protocol. He had not reported it.
87 percent was a number in a known range. It did not require interpretation. He was bringing it to the meeting as a number, not as evidence.
The dormant-baseline monitoring protocol had been designed for exactly this problem: the trace index moved, and if you watched it closely enough, you began to attribute meaning to the movement. The protocol specified a designated check schedule and a documentation format that emphasized recording over interpretation. You checked the number. You recorded it. You did not open the contextual weight assessment unless the trace index crossed the threshold specified in the monitoring agreement. Branch 12's threshold was 94 percent. It had never crossed 94 percent. Marcus had checked forty-three times, and the number had never exceeded 91 percent, which it had reached for two consecutive months in 2046 before returning to its baseline band.
The novel's appointment-book man had spent eleven chapters trying to consolidate his calendars into one system. He had used three different methodologies, hired a consultant, built a spreadsheet, and finally hand-transcribed everything into a single book — and then in chapter ten had discovered that the single book lost something he could not name, some quality the separate books had carried through their interaction with each other, and he had spent chapters eleven and twelve opening the original books back up one by one. By the end, five were open. He had stopped consolidating. The novel ended.
Marcus was not sure the novel was about appointment books.
He was not sure what the September 17 meeting would recommend. The proposal for suspension had an elegant internal logic: if the purpose of the dormant-baseline protocol was to accumulate uninterpreted data until the interpretive advisory group had enough baseline to work with, and if forty-three months was sufficient, then the protocol had served its purpose and could be retired. The data was archived. The automated flagging system could handle threshold crossings. The human observer function was no longer necessary.
He could see no flaw in this logic. He was also aware that the logic assumed the purpose of the dormant-baseline protocol was what the documentation said it was, and that forty-three months of performing a function gave you some relationship to that function that the documentation might not fully account for. The appointment-book man's consultant had told him, in chapter four, that the problem with his calendar system was inefficiency. The consultant had been right. The consultant had also missed something.
Marcus did not think he was indispensable. He was one person who had checked a number forty-three times and written it down. The system had been running without his intervention for the entire forty-three months. Branch 12 had not needed him to reach 87 percent. Whether his presence had done anything the system would miss was a question he could not answer from inside the function.
He was also aware that Annika had stopped answering twenty-six days ago, which was forty-three days after the last Branch 12 reading he had flagged as worth a secondary note — not a threshold crossing, just a pattern shift worth documenting. He had documented it. He had not mentioned the timing to anyone. He was not drawing conclusions. He was a dormant-baseline observer. He was not drawing conclusions.
He finished his coffee and made a second cup.
At 7:14 AM he opened his forty-three month notebook to the last entry: September 15, trace index 87 percent, no threshold crossing, no contextual flag triggered, session closed. He read it. He did not add anything. The entry was the entry. He closed the notebook.
He dressed and ate breakfast. The transit system had completed its morning reconfiguration by 6:45. The delivery systems would finish their first pass by 8:30. He knew this without checking anything; three and a half years of early mornings had given him the city's operating rhythm as ambient knowledge, the same way the dormant-baseline monitoring had given him the Branch 12 trace cycle.
He was not afraid of what the meeting would say. He had considered this on the walk to the window: whether he was afraid. He did not think he was. The function had been assigned and he had performed it and now the function might be over. That was a plausible outcome. He had not accumulated enough of himself in the function to be frightened of its ending. Or if he had, the accumulation was not registering as fear.
It was registering as something. He noticed this without naming it. He was a dormant-baseline observer. He did not name things.
The Branch 12 dashboard was visible on his secondary screen, dormant. The screen showed the last session data: 87 percent, September 15, 14:23. The contextual weight assessment bar was grey. He had not triggered it.
At 9:22 he picked up his bag and the forty-three month notebook and left the apartment.
The Branch 12 dashboard, still open on the secondary screen, continued to show 87 percent. The contextual weight assessment bar was grey. The system did not require his presence to continue. It had not required his presence for forty-three months. What his presence had provided was something the system's documentation called baseline stability and what Marcus, if he were annotating this rather than merely recording it, might have called a witness who did not yet need to know what he was witnessing.
The meeting would tell him whether this had been useful.
The trace density measure had been introduced in the system's third iteration, four years before Marcus had taken the monitoring assignment. It tracked something the documentation described as the secondary attribution layer: the implicit weight-assignment events that preceded conscious attribution decisions, measurable in the system's cognitive index but occurring below the threshold of deliberate reasoning. The system's architects had argued that the secondary layer was where behavioral pattern shifts originated, and that by the time the shift was visible in the primary attribution layer, it was already weeks old in the secondary. Branch 12 was the monitor for one participant cluster whose secondary attribution patterns had been stable for long enough that the system had placed it in the dormant category: not static, not inactive, but not requiring active interpretive engagement. Just watching.
Marcus had been watching.
He understood, at 6:12 AM, that the proposal for suspension was not about whether Branch 12 was worth watching. The secondary attribution layer was real. The trace density measure was reliable. The forty-three months of data were in the archive and would remain there whether or not he continued to generate entries. What the proposal was about was whether a human observer — specifically him, specifically at the designated check cadence — was adding something to the archive that an automated threshold monitor could not. The interpretive advisory group had evidently concluded the answer was no. Marcus had been asked to attend the meeting and present his monitoring records. He had not been asked for his opinion.
He was also not sure he had one.
The appointment-book man had not been asked for his opinion about consolidation either. The consultant had presented the analysis and the recommendation and the man had tried to implement it. The difference was that the appointment-book man had eventually stopped implementing it and the novel had ended there, with the books still open. Marcus did not know what the equivalent would be for him. He did not know what he would do if the meeting recommended suspension and he disagreed. He did not know if he disagreed. He thought he might be about to find out.
He put on his jacket. The notebook went into the bag. The novel stayed on the table, five pages open, which was not how he had left it — he had put it down with the last page face-down, not face-up — and then he remembered that he had opened it again at 2 AM to read the ending a third time, and had left it open. He did not close it before leaving.
He did not check the dashboard before closing the door.