The Tier 4 district batch had arrived Tuesday at 8:17 AM from the northern assessment node, and Lived-Cal had processed it in eleven minutes.
Yoon Gyeol-ri ran it through her personal filter over the course of four hours, between her other assignments, returning to it in the gaps. This was not official calibration protocol. Official protocol was Lived-Cal assessment, anomaly flagging, comment field, submit. She did all of that. She also did the four hours.
The calibration system had given the batch a 94.
Yoon Gyeol-ri had given it a 71.
These two numbers were not in conflict, technically. They were measuring different things. The calibration system — Lived-Cal, the Fidelity Assessment Engine that had been standard in The Seam since the FidelBase rollout in 2030 — measured signal integrity against the benchmark registry: spectral consistency, compression artifact detection, temporal coherence, emotional marker amplitude within expected parameters for the demographic cohort. At 94, the batch was in the top tier. Near-perfect fidelity. The memories it contained were, by every standard the system could apply, complete.
Gyeol-ri's 71 was not a fidelity score. The system had no name for what it was.
She had developed the personal calibration filter over eighteen months of calibration work in The Seam, alongside the kkaeji practice and not entirely separate from it. The two had started at different times — calibration work first, kkaeji later, when the neighbor across the hall from her apartment had taken her to the first tteum-jib exhibition she attended, in 2033, a set of twelve pieces made from unprocessed sonic memory and the institutional forms that had rejected them. She had stood in front of one of the pieces for a long time. It was a form — a standard Lived-Cal anomaly report — with every category field left blank and the comment field filled to capacity with a single sentence, repeated in smaller and smaller type until it was illegible. She had not been able to read the sentence. She had not needed to. The piece was about the comment field. She understood this immediately. The filter began as a note she kept beside her calibration terminal — a list of things she noticed that Lived-Cal did not flag, things that were present in the recordings but did not register in the standard metrics. The pauses, primarily. Emotional cadence. The relationship between what a person's body was doing in the moment of recall and what the recall presented as the affective content of the memory. Lived-Cal measured the emotional markers; her filter measured something about how they arrived.
The calibration work and the kkaeji practice had been running in parallel for two years before she understood they were the same inquiry. The calibration work asked: what does Lived-Cal miss? The kkaeji practice asked: what does the miss look like, as a thing you can stand in front of? The filter had emerged from calibration work, but the filter's purpose was kkaeji. It was a tool for making visible what the system classified as acceptable range.
She had never submitted the filter to anyone. She had not been asked to. The calibration team processed batches through Lived-Cal and produced fidelity reports, and the fidelity reports went to the receiving parties — the memory archives, the fidelity insurance assessors, the legal teams handling contested recalls — and the standard was the standard. Her personal calibration was what she did with the recordings before she passed them up the chain, in the forty minutes between when a batch arrived and when it needed to be logged. Forty minutes of looking at something Lived-Cal could not see.
The Tier 4 district batch had arrived Tuesday and she had given it four hours she did not officially have.
The piece she made from the batch — "Interview Posture (Tier 4)" — was four silences arranged on a grid. She had not shown it to anyone. She had printed it and pinned it to the inside wall of her locker, which was the closest thing she had to a private space in the calibration center. The wall of her locker was not regulated; the calibration center's monitoring system covered work surfaces and network activity, not lockers. She was not sure why she cared about this, since "Interview Posture (Tier 4)" was not secret — it contained nothing that was not in the batch report, which was itself not classified. Four millisecond counts on a grid. Anyone could have derived them from the report.
She cared because the piece was hers in a way the report was not. The report was a fidelity assessment. The piece was a question: if the pauses are too even, who taught them to pause?
She did not know the answer. She was not sure the piece was supposed to answer it.
The second piece came from the same batch: the same recording rendered in two versions, side by side. The Lived-Cal score in the first column: 94. Her personal calibration score in the second: 71. No other text. No explanation of the scoring methodology. The gap between the numbers was the piece.
A colleague had asked her what she was working on outside the office. She had said: a series of silences. The colleague had waited for more, and she had not added more, and the colleague had returned to their batch. Gyeol-ri had thought about this afterward: the accurate description of kkaeji work was often indistinguishable from a non-answer. She had been doing kkaeji since 2033 and she still could not explain it in a way that sounded like an explanation. She had pieces in the tteum-jib gallery on Seam Row — three of them, each from a different batch over the last two years — and when people asked her what they were about, she said: the gap between what the system sees and what is there. This satisfied some people and confused others. The ones it confused were usually the ones most oriented toward the standard fidelity metrics, who thought a gap in measurement was a problem to be solved rather than a subject for art.
She did not think they were wrong. The gap was a problem to be solved. She was also making art from it. Both things were happening.
She submitted the batch report at end of day: fidelity scores, technical notes, flagged anomalies. The anomalies from the Tier 4 batch she had marked as "affective cadence irregularity, within acceptable range" — because they were within acceptable range, technically, and because she had no official category for "too even to be uncoached." She had been doing this for three years. Marking things within acceptable range because the thing she was actually noticing was not in the category system.
The category system had been developed by a committee in 2029 to standardize fidelity reporting across Lived's nine assessment centers. It had twenty-three anomaly categories and a comment field. She used the comment field. Nobody had ever asked her about a comment field entry.
She walked home along the lower edge of The Seam at 6:40, after the calibration center had emptied out and the overnight batch queue had been handed off to the automated intake system. The sky over The Seam had gone the particular grey-violet of early evening in the district — the light absorbed by the close-built towers, the certification kiosk displays beginning to glow.
She passed the tteum-jib galleries with their diffuse window light, past the certification kiosks where people paid for their Lived-Cal assessments, past a group of three talking outside one of the district's tier-recalibration offices — people whose certified fidelity level had been reassigned downward after the last audit cycle, who were now navigating a city in which their memories were officially less reliable than they had been six months ago. This happened to a few hundred people a year across The Seam. The tier-recalibration offices were one of the institutions the kkaeji movement had made its subject from the beginning: the bureaucratic processing of memory's legitimacy.
She passed the tier-recalibration office and its small cluster of people and thought, as she often did when she walked past it, that she wanted to make a piece about what it felt like to have your memories officially reassigned downward. Not the bureaucratic process — the galleries had covered that. The phenomenology of it: waking up the morning after the letter arrived and trying to remember something, and knowing that the memory itself had not changed, but that its official fidelity rating had. Whether the knowing changed the experience of remembering. Whether you could trust a 0.84-certified memory the same way you had trusted it when it was 0.91.
She did not have that piece yet. She was not ready for it. It required a kind of access to someone's interior she did not have standing to claim from the outside. Maybe it required being inside it herself. She did not know.
She did not stop. She had a new piece forming and she wanted to get home before she lost the shape of it.
The shape was this: not the gap between the two scores — she had already made that piece. The next piece was the gap between the comment field and the anomaly categories. Between what the system had room to hold and what it could not. Not the fact of the gap, which she had established. The texture of the gap, from inside it, after three years of working in it.
She got home and opened her notebook — a calibration notebook, the official kind, with pages printed in the diagnostic grid used for spectral analysis. She had been using the wrong side of the pages for years: the blank back, unlined, unofficially hers. The calibration center issued them by the box. No one had ever asked her how many she used.
She wrote down the shape before it left her. Not the piece yet. The question the piece would hold.
She also wrote three lines below the main shape: the tier-recalibration piece is not mine to make yet. I do not know enough about what it feels like from inside it to make it honestly. This is a legitimate reason to wait. It is also the kind of thing I could say forever and use it to stay safe. Watch the difference.
She closed the notebook. The notebook was the official calibration kind — the wrong side of official pages, the blank backs, unofficially hers — and she kept three of them in the drawer beside her bed and had never once thought about what would happen to them if she left the calibration center. They contained nothing formally significant. They contained everything she had noticed for three years that was not in the category system.
She made tea and sat with the shape of the next piece, which was not yet a piece. It was the thing that arrived before the piece, which she had no name for in Lived's category system and several names for in the kkaeji community: the pre-form, the negative space, the gap before the frame. The system could not see it. She could. That was the whole practice.
The Lived-Cal score said: 94. Her score said: 71. The batch was from the Tier 4 district. The pauses were too even. She did not know who had taught them to pause that way, and neither did the system, and the difference between those two versions of not-knowing was what she was going to make next.
She wrote for an hour, which was longer than she usually wrote in the pre-form stage, and then she stopped because she had what she needed and because writing past what you needed was how you used up the thing before you could make it. She made tea. She looked out the window at The Seam's evening — a few lights in the towers across the narrow street, the certification kiosk at the corner ticking through its idle display, cycling through fidelity tier examples the way it did every hour. 0.90, 0.85, 0.80. The numbers that decided how much of your remembered life you were officially allowed to trust.
She drank the tea. Tomorrow there would be a new batch.