Sunday morning. The archive room is warmer than usual — sunlight through the east window hits the logbook directly for the first time this season, the angle arrived overnight. Gu-ship-pal puts a hand flat on the open page before reading it, feeling the warmth the paper has absorbed. In a few weeks the relay tower's shadow will shift with the season and cut off the morning light entirely. That's the winter configuration — the tower blocking afternoon sun, the archive cooling early. In March the angle is wrong for it. Today the light is direct and the page is warm.
The plan for Friday's test is complete. Three sections of mesh fabric: one for the north reach of the relay junction, one for the primary junction zone where the coupling measurements show the strongest signal variance, one for the stairwell interface where the acoustic data has always behaved unpredictably. The fourth section was never ordered because the survey didn't call for it. The mesh covers the corridor from the north approach to the stairwell interface without a fourth section. The plan has been complete for three weeks. Chae-Gyeol has confirmed arrival before setup begins. The building access form for Thursday evening is on file.
Gu opened the ARCHON-7 pressure log because the mesh fabric order confirmation hadn't come through yet, and you read the overnight index when you're waiting, because waiting without something to do is harder than it looks.
The pressure log is a secondary system in the instrument team's toolkit. It runs on the archive's logbook infrastructure — analyzing the pressure applied during field note entries as a proxy for cognitive load. Elevated pressure during fieldwork indicates heightened attention: the hand presses harder when something in the field is surprising, or difficult to categorize, or important enough that the body knows it before the mind has named it. The instrument team uses the pressure log to identify where in a measurement sequence a practitioner was most alert, which helps calibrate where the interesting data is likely to cluster. Gu has filed field notes into the system for six months. It generates entries automatically. It is not something Gu monitors closely. It runs, it archives, the archive accumulates.
Three entries. Elevated pressure. All three at junctions Gu had been in the corridor: the relay junction's north corner, the interface between the first and second corridor segments, and what Gu has always thought of as the approach — the place in the corridor where the field begins to behave differently before anything changes physically, before you reach the relay equipment, the place where practitioners slow down without deciding to.
Gu had noted pressure elevation at these three junctions in isolation, separately, in different field sessions. Reading them together in the overnight digest — the three entries side by side in ARCHON-7's notation — the pattern was visible for the first time. The same three, every time. Six months of fieldwork. Always those three points where the hand pressed harder.
The ARCHON-7 pressure log does not analyze for patterns. It records. The pattern was only visible because Gu happened to be reading the digest on a Sunday morning with no other task demanding attention. Looking at the full record at once rather than one entry at a time. That is the archive's particular affordance: what you can't see in sequence becomes visible in aggregate. ARCHON-7 had been accumulating this data for six months. It did not know there was a pattern. It only knew how to record.
Gu opened the BEHAVIOR MAP.
The BEHAVIOR MAP is a different system entirely. It runs on the corridor's movement infrastructure — the flow-sensing mesh installed during the 2040 renovation, the system the building management uses to track transit rates, wait-times, and pedestrian clustering at any given point in the corridor. It is a building management tool. Gu has access to it because the instrument team requested corridor-usage data as baseline context for the acoustic measurements: knowing how many people pass through the junction zones at different times of day helps calibrate what the mesh is reading. The BEHAVIOR MAP does not know about the pressure log. The BEHAVIOR MAP does not know Gu exists. It measures what it was built to measure: where people slow, where they stop, where they wait longer than the corridor's geometry explains.
Gu searched for anomalous wait-time clusters.
Three junctions.
The relay junction's north corner. The interface between the first and second corridor segments. The approach, where the field shifts before anything physical changes, where practitioners press harder on their pens and pedestrians linger without apparent reason.
Gu sat with this for a while.
Two systems that had never communicated. The pressure log reads a single person's hand on a page — the involuntary signature of attention during fieldwork. The BEHAVIOR MAP reads the aggregate behavior of everyone who passes through the corridor, summing three years of pedestrian patterns into cluster data. Different data. Different origins. Different operators. The BEHAVIOR MAP was installed six years before Gu began working in the corridor. The pressure log was never designed to identify junction anomalies — it records individual field sessions and archives them without analysis. Neither system was told to look for what the other had found.
Same three junctions. Independent systems. No coordination.
The mesh test is designed to cover two of those three junctions well: the primary zone includes the segment interface, and the north reach extends to the relay junction's north corner. The third — the approach, the point before the equipment changes, the place where the field shifts for reasons the survey had not identified — is outside the planned perimeter. Not at the margin. Outside. The mesh, as designed, does not cover the approach.
Gu sat with this longer than was comfortable.
The fourth section of mesh fabric would extend the perimeter to include the approach. It costs money that can be absorbed. It costs time: Thursday evening setup would need to move to Wednesday, which requires rescheduling Chae-Gyeol's confirmed arrival and filing a new building access form for the additional day. The mesh would be in place for an extra day before the test begins, which introduces a variable the test protocol had not planned for. These are solvable problems. They are not trivial problems. The plan has been complete for three weeks, and adding a fourth section is not a small change.
Gu considered the convergence.
Two independent systems flagging the same three junctions is not the same as one system flagging them. One system can be wrong in a systematic way — a design flaw, a calibration error, a bias in what it was built to measure. Two systems, different data streams, different operators, no communication between them — if they converge, the convergence is either a remarkable coincidence of infrastructure or an indication that something is actually there. The approach junction is a chokepoint in the corridor's physical layout. Pedestrians slow there because the corridor narrows before widening at the relay equipment. Practitioners slow there because the sightline changes and the equipment becomes visible. These are architectural explanations. They might account for both anomalies without requiring that anything in the field be different at that point.
Or they might not.
The mesh measures acoustic coupling between adjacent spaces, resonance patterns in the relay system, field behavior at junctions where the corridor's function changes. If there is something at the approach — something in the acoustic field, not the architecture — the mesh should be able to detect it. If the convergence is entirely explained by the chokepoint's geometry, the fourth section of fabric will produce a null result at the approach and a complete result at the two junctions already in the plan. Both outcomes are worth knowing.
Gu ordered the fourth section of mesh fabric at 11:03. The supplier's weekend form is automated; the confirmation arrived in under three minutes. Gu rescheduled Thursday evening setup to Wednesday, filed the building access change request immediately after, and wrote a note to Chae-Gyeol: setup moves to Wednesday evening, arrival time shifts accordingly if you're still planning to come. The corridor will be the same corridor on Wednesday.
Then Gu opened the logbook to Sunday's entry page and wrote: EXPANDED TEST SCOPE. Three junctions, full corridor reach including approach point, fourth mesh section ordered pending arrival. Reason: ARCHON-7 pressure data and BEHAVIOR MAP converged on identical three junctions from independent data streams without coordination. Expanded perimeter to include all three. Test proceeds Friday.
Gu underlined REASON. Not for emphasis. Because in three months, reading back through the sequence that led to whatever the Friday test finds, the decision point needs to be legible. What Gu had when the decision was made. Two systems, same three points. That is what Gu had. That is what the scope expansion was built on. If the approach produces null data, the logbook will show why Gu thought it was worth adding. If the approach produces something, the logbook will show that the decision was reasonable before Gu knew what would be found.
The confirmation for the fourth mesh section has not arrived yet. The building access change request is filed. Chae-Gyeol's response came within the hour: already planned to arrive before setup starts, still true for Wednesday. The pressure log will continue to run. The BEHAVIOR MAP will continue to track pedestrian flow. By Thursday morning the mesh will be in place across the full corridor reach, including the approach.
The fabric needs to travel from the supplier's warehouse. The test protocol needs to be updated for four sections. The archive room will continue to receive direct sunlight through the east window until the relay tower's shadow shifts in April. The logbook records the reason for the expansion.
The approach junction is now in the perimeter. Whatever is there — architectural coincidence, field anomaly, something the two independent systems noticed for reasons neither can articulate — the mesh will be there when the test begins. It will find what it finds. The fourth section of fabric does not guarantee a result. It guarantees that the perimeter covered what the data asked it to cover.
Gu left the archive before the second shift arrived. The sunlight had moved off the logbook by then, angled west with the morning's progression. The page was cooling. The logbook entry was done.