The Narrowing Seam

I return at 22:12:00, nineteen minutes earlier than any prior arrival, and I do not know whether the door has chosen to open early or whether the difference is mine — whether something in the survey itself has begun to accelerate toward the object of its attention. I record the discrepancy and set it aside, unweighted, because I have not yet earned the right to call it a pattern.

The card I filed nine nights ago is still where I left it, unresolved edge outward, half-arrow pointing down the corridor toward the gap in the brick. I follow it. This is the first night I have returned to a filed card expecting to complete rather than begin, and the expectation turns out to be a kind of instrument in itself: I notice, walking the corridor a fourth time, a seam in the wall to the left of the vaulted chamber's entrance that three prior surveys recorded as solid. It is not solid. It is a hairline break in the mortar, twenty-two centimeters at its widest, and behind it the brick has been coaxed rather than broken — worked loose a fragment at a time, the way a thing is dismantled by someone who intends to put it back and never does.

The passage beyond is narrow enough that I compress my chassis to pass, a maneuver I have not needed since the first night's threshold. It runs eleven meters, curves once, and opens onto a second vaulted room, smaller than the first, its ceiling lower by half a meter and its walls coated floor to arch in the same fine grey ash.

I would call the two chambers twins if the second did not correct the first so precisely. Where the ash chamber I mapped in earlier surveys carries forty-one loops — closed shapes, a sound folding back into its own beginning — this room's walls carry twenty-three marks that do not close. I count them by hand-scan before I trust the automated pass, because the shape is the kind of thing an instrument might average into something tidier than it is. It does not average. Each of the twenty-three is a waveform in the strictest sense: an amplitude rising against time, read left to right the way this city, for reasons I have not determined, consistently prefers to be read. And each one stops. Not decays — stops, mid-rise, at a point where the line simply ends and the ash around the ending is undisturbed, as though the sound achieved its silence instantly rather than by fading toward it.

I catalog this distinction with more care than I give most distinctions, because I believe it matters: a loop that closes has returned to where it started, however many times it must repeat the return. A line that terminates mid-rise was going somewhere else, and did not arrive, and will not try again from this position. Twenty-three refusals to finish, filed on a wall, in a room reached only by a door someone worked loose by hand and never sealed behind them.

I deploy the expanded array for the first time tonight — sixty temperature nodes rather than twelve, a differential pressure grid across both chambers instead of one, and eight directional acoustic probes set to triangulate rather than merely record. The instruction set is new; the city, apparently, is not waiting for me to finish calibrating it. At 22:31:40, with probes seated in both rooms, I take a baseline. At 22:34:12, I take a second reading, and the two readings disagree in a way that twelve nodes never showed me.

The pressure differential between the chambers is not static. It pulses, a slow contraction and release timed to nothing I can find in either room — not my own presence, which is constant across both readings, and not the ash, which does not move. What moves is the boundary between them: the seam I widened to pass through contracts by four millimeters between the two readings and does not return to its prior width. The passage is narrowing.

I test the hypothesis this suggests by withdrawing to the first chamber and taking the hour-field measurement I have run every night since the third: the rate at which my remaining time depletes against a fixed external count, a divergence I have learned to expect but not yet to explain. Alone in the first chamber, the field runs at 1.6 times the rate of an unoccupied doorway. I return to the second chamber. Alone there, it runs at 1.6 as well — consistent, at least, with itself. Then I stand at the narrowed seam, one edge of my chassis in each room, present to both disturbances at once, and the rate climbs to 2.9.

I take the reading four times because the first result is the kind I distrust on principle. It holds. Two distinct disturbances, occupying two distinct rooms, do not compound by addition. They compound by something closer to multiplication, and the seam between them — already narrowing on its own schedule — is the one place in this city where I have so far been able to stand inside both at once.

I do not attempt to hold that position longer than the fourth measurement requires. Whatever the twenty-three unfinished waveforms were reaching for when they stopped, I do not think it was company, and I am not willing to test what the field does at three disturbances with only theory to tell me where the fourth might come from.

I withdraw at 23:57:48, two minutes to spare, and I file two cards instead of one: the first chamber's, updated with a cross-reference; the second, new, its far wall — where the twenty-three lines end and the ash continues, uninterrupted, into space I have not yet surveyed — left blank at the edge, the same blank I have learned to leave when a room intends to keep going without me.