Yoon Gyeol-ri arrived at the Boundary Layer micro-residency at half past seven on a Thursday evening in mid-autumn, following directions from a friend who had described it as a basement that takes itself seriously. The friend was not wrong.
The entrance required a fidelity-check: a wall terminal near the door, brushed aluminum, the Distributed Recall network logo in the top corner. She pressed her handheld to it. The terminal returned her certified-tier in green — the number was 94.7, slightly better than last year's annual pull. She had never known what to do with that number. It sat in her calibration record like a grade she had not studied for. The terminal also flagged that the event itself was running under a non-standard certification protocol — four independent streams, separately certified, no composite. She had never seen that flag before. In three years of attending Distributed Recall-affiliated exhibitions in the Seam district, she had never encountered a non-aggregate filing. She accepted the flag, noted it in her field notes as entry point: protocol anomaly, and went in.
The space held twelve people. Folding chairs arranged in a loose semicircle. A woman near the back was typing notes on a translucent tablet; the light from her screen made her face greenish and intent. The walls were raw — the kind of basement that had stopped apologizing for being a basement. One wall was shimmer-panel, certified, probably original to the building's 2027 renovation. The other three were uncertified concrete, predating the Fidelity Acts by twenty years. The seam between them ran at eye level, a clean horizontal line where two different administrative realities shared the same load-bearing function.
The work was a reconstruction: a single afternoon from a Distributed Recall donor — a woman in her forties, a harbor worker, an afternoon in 2029 when she had helped carry a neighbor's water filtration tank up four flights of stairs after the building pump failed. The afternoon ran forty-seven minutes, told across four screens that did not synchronize.
Gyeol-ri watched the left screen for seven minutes. She watched it the way she watched everything: carefully, looking for the thesis. She was looking for the primary screen. She was trying to find which was the anchor from which the others deviated.
She found no anchor. There was no primary.
The first screen showed the climb: the tank, the stairs, the woman's shoulders, the neighbor's breathing. The second screen showed the harbor — not that afternoon, but the same woman three weeks earlier, a day the donor had flagged in her consent notes as the last good day before the strike. The third screen showed the filtration tank itself: its manufacture date, its certification level, its projected lifespan, its recall-index as assigned by the municipal water authority — a number that had changed three times in the past year as the authority updated its fidelity scoring rubric. The fourth showed the stairwell after — empty, a long shot, nothing moving, just the ambient sound of a building that did not know it had been carrying something important an hour before.
Each screen was accurate to its own timeline. Each screen had been certified by the Distributed Recall network. Each screen was, in the network's formal language, fidelity-verified at submission. Together they did not add up to a continuous experience. They added up to something else: a question about what continuity costs and who decides.
Gyeol-ri opened her field notes on her handheld.
She had been writing notes about fidelity for three months. She had six pages on the closed gallery, the Fidelity Board's 91.4 score, the calibrators' readout, the measurement apparatus. She had been writing as though fidelity were a quality of the object — of the work, of the memory, of the reconstruction. A number you could assign.
She wrote: first time observing work where the frame is the argument.
The Board had issued a 91.4 on the closed gallery. The Board had watched the work. She did not know which screen. She did not know if the Board had watched all four or one or had reviewed the certification records and assigned a composite score from the metadata. The number 91.4 was precise. The number 91.4 was specific. The number 91.4 was, in the Fidelity Board's framework, a measurement.
She added a line: the Board was watching one screen.
The work ended without announcement — the screens simply went dark at their own times, the first screen at minute forty-seven, the fourth screen at minute forty-nine, the second and third somewhere between. The twelve people in the room did not applaud. They stood for a moment in the dark and then the lights came up and someone near the back said: how is it calibrated?
The artist said she had submitted each stream separately and received four separate fidelity scores. She had averaged them. She said this without apology and without explanation, as though this were the obvious method. Someone asked if averaging was the right aggregate. The artist said she did not think there was a right aggregate. She said she thought that was the piece.
The woman with the translucent tablet raised her hand. She asked whether the work had been through peer-review certification or direct-submission. The artist said direct-submission — she had filed the streams through the standard Distributed Recall portal, used her own agent-ID, waived the peer-review queue. The woman with the tablet said: so it's got DR provenance, but not Board-endorsed. The artist said yes. The woman said: does that matter to you? The artist said it had not, until someone asked.
Gyeol-ri wrote: the aggregate is not a solution. it is a position.
She had three months of field notes. She had a thesis that was forming into something she would have to write down formally in four months for the kkaeji exhibition committee. She had been worried, for three months, that the thesis was not hers — that the closed gallery had done the thinking and she was just reporting it. She had come tonight to find out if a second site would confirm or complicate what she had found in the first gallery. She had thought: if two different fidelity-measuring systems produce the same failure mode, then the failure mode belongs to the system. If they produce different failure modes, then the problem is more specific and she does not yet understand it.
The Distributed Recall piece had a different failure mode than the closed gallery. The closed gallery's failure had been about apparatus — calibrators reading the work instead of watching it, measurement replacing encounter. The Distributed Recall piece's failure was about governance: a system designed to certify individual streams, applied to multi-stream work, producing a composite that no one had authorized. The systems were different. The structure of the failure was the same. The measurement had answered a question the work had not asked.
Afterward, in the narrow corridor between the gallery space and the street stairs, Gyeol-ri found herself behind the artist in the queue for the exit terminal — another fidelity-check, this one logging attendance for the venue's Distributed Recall compliance record. The terminal asked whether she wished to register as a witnessed presence for the event's certification archive. She pressed yes without thinking. She had always pressed yes. It had never occurred to her not to.
The artist was still in front of her, speaking to someone — a curator, maybe, or a funder. Gyeol-ri heard her say: the donor wanted all four streams certified separately. She said she didn't want the network averaging without her permission.
Gyeol-ri stopped walking.
The donor had thought about this. The donor had made a formal choice — had navigated the Distributed Recall consent interface, which Gyeol-ri knew from her own calibration training was not simple, and had filed a non-aggregate consent clause. The donor had decided that her afternoon was not a thing that could be averaged. She had decided this and filed it and the network had honored it and the artist had built the piece around it.
The artist had not invented the four-screen form. The donor had.
Gyeol-ri stood still in the corridor for long enough that the person behind her touched her shoulder gently. She moved forward. She registered her presence. The terminal returned a green acknowledgment and appended her attendance to the event's certification record in the Distributed Recall archive, where it would remain for thirty years under the standard preservation tier.
She went up the stairs.
Outside, the Seam corridor ran east along the transit line, the buildings on the left certified all the way up — shimmer-panel facades, the soft luminescence of authenticated materials in low light — and the buildings on the right predating the Fidelity Acts, matte concrete, no shimmer, nothing to authenticate. Both sides were load-bearing. Both sides were real. The distinction was not structural; it was administrative.
She had commuted this route for two years. She had calibrated the left buildings in her head without calibrating them — she knew which were certified, which were uncertified, the same way she knew which cafés had Distributed Recall compliance signs and which did not. She had stopped seeing the seam.
On the train, she opened her field notes and added to the thesis section:
The donor chose non-aggregate consent. The artist honored it. The work is the donor's formal claim about the nature of her afternoon: that it cannot be averaged without falsifying it. The Fidelity Board's 91.4 on the closed gallery is a response to a claim the Board never asked about. The Board averaged. The donor said don't. The Board said 91.4 anyway.
This is not a controversy. The Board is not wrong by its own rules. But the donor is also not wrong. The same afternoon, certified at different levels depending on which set of rules you apply. The measurement is not neutral. The measurement is a position.
She stopped. Thought about the six pages she had written about the closed gallery. She had been building toward the idea that the 91.4 was a measurement of the gap between the apparatus and the field — that the calibrators had been reading one screen while the work insisted on four. That thesis was true. It was also smaller than she had thought.
Below that, in smaller type, she added a line she would not put in the academic version:
My thesis does not belong to the first gallery. It does not belong to me. It belongs to the donor, who figured it out before either of us, and filed the paperwork.
The train passed the certification boundary between the Seam district and the Inner Fidelity corridor, where the infrastructure was fully authenticated and the transit system itself was certified — every arrival and departure logged in the municipal Distributed Recall record as a matter of urban continuity compliance. Gyeol-ri had ridden through this boundary twice a day for two years. She had never thought about what it meant that the train's arrival was being remembered on her behalf.
She had 94.7. She had never known what to do with that number.
She looked at the certification boundary marker through the window — a small blue indicator at the platform edge, standard municipal notation, readable only if you knew what to look for. She had learned to read them in her first year at the tteum-jib gallery, when her supervisor had explained that the Fidelity Board's jurisdiction ended at the Seam and the Distributed Recall network's jurisdiction extended city-wide but at different resolution. She had written it in her notes at the time: two systems, same city, different granularity. She had thought it was administrative detail. It was the whole thing.
The train came to her stop. She got off. She did not close her field notes. On the platform, under the blue certification boundary marker, she stood for a moment in the transit noise and the cold. Then she walked home.