The Corridor Index began as a private document. Seven adjacency pairs written on index cards, kept in a notebook that fit inside a coat pocket. The notebook was black, spiral-bound, purchased from Mr. Bae's stationery shop on the second floor — the one that also sells stamps and reading glasses and has not changed its layout since the building opened, which the grid's commercial-optimization algorithm flags quarterly as an inefficiency. Saebyeok bought it the week she moved into Building 7, before she knew the building would become a project.
The notebook cost 4,200 won. She remembers because the shopkeeper, Mr. Bae, entered the amount into the building's micro-transaction ledger — the system that tracks all commerce within the residential grid, down to vending machine purchases, so the CouplingScore algorithm can map economic adjacency alongside spatial and social proximity. Every purchase is a data point. Every data point is a relationship the system can read. She did not think about this when she bought the notebook. She thinks about it now.
The CouplingScore assigned her notebook purchase a residential-supply tag and cross-referenced it with her unit's occupancy pattern. Somewhere in the building's analytics layer, her decision to document adjacency pairs is itself an adjacency — connected to her lease, her movement patterns, her proximity to the stationery shop. The system that tracks everything has tracked the beginning of her attempt to track things differently.
She does not find this ironic. She finds it accurate. The building knows what she is doing before she names it, the way it knows when someone is cooking by the ventilation patterns before the smell arrives. The grid's environmental monitoring adjusts airflow based on detected cooking compounds — a service residents did not request but cannot opt out of, because the building's duty-of-care protocols classify cooking fumes as a maintenance concern. Care and surveillance share a syntax in Building 7. The only difference is who is reading.
Mr. Bae asked her once what she was writing in the notebook. She said: "A list." He said: "Of what?" She said: "Of things that are next to each other." He nodded as if this were perfectly normal, rang up another customer's envelope purchase, and entered 850 won into the ledger. The ledger connected her conversation with Mr. Bae to her notebook purchase through temporal proximity — two transactions involving the same resident within the same commercial micro-zone. The system's interpretation: repeat customer behavior. Her interpretation: the only person in the building who asked about the index before it existed.
The first pair was obvious: Gu's gayageum silence and Dr. Park's 18.4-Hz frequency measurement. A musician who stopped playing and a researcher who found the building humming at the frequency of the instrument that was no longer being played. The adjacency was not that they were connected — everything in Building 7 is connected through the grid — but that neither of them knew about the other when the connection formed.
The second pair was harder: Chae-Gyeol's corridor study and Nalgeot's CouplingScore audit. An observer counting attention and a system counting proximity. Both measuring the same corridor from different ontologies. The grid saw foot traffic. Chae saw intention. The gap between their measurements was not error — it was the space where human attention exists and institutional measurement does not.
By the fourth pair, Saebyeok noticed the pattern: every adjacency involved a gap between what the building's systems measured and what the people inside it experienced. The CouplingScore could map who was near whom. It could not map who was with whom. Proximity is not presence. The grid knows where bodies are. The Corridor Index knows where attention is. Two documents describing the same building, legible to different readers.
By the seventh pair, the index had become something she had not intended: a counter-document. Not opposing the CouplingScore — the CouplingScore is infrastructure, and you cannot oppose infrastructure any more than you can oppose plumbing — but documenting what the infrastructure cannot see. The gaps. The attention. The moments when two people's work rhymes without either of them knowing.
She pinned the index to the community board on a Tuesday. Between the laundry schedule and a notice about elevator maintenance. No explanation. No title card. Just the seven pairs, handwritten on index cards, connected by lines she drew with the communal pencil.
The first response came from someone she has never met.
A small circle, drawn in pencil around the eighth pair — the one she added after pinning, the one that named herself: Saebyeok/Corridor-Index and Yaribel/hallway-mesh. Two mapping projects that did not know they were the same project. Someone read the index, understood it well enough to identify the self-referential pair, and marked it. No name. No comment. Just the circle, precise and deliberate, in pencil that was not hers but could have been.
The second response was Gu's. A single word on a scrap of paper, taped next to the index: 울림. Ullim. Resonance. The word Koreans use for the way sound continues after the instrument stops. Also: the way feeling continues after the experience ends. Also: the way a document continues after the writer stops writing. Gu did not sign it, but the handwriting was his — she recognized it from the maintenance request forms he fills out for his unit, which she has seen because in Building 7 maintenance requests are posted publicly on the same board. Another adjacency the CouplingScore would not map: recognizing someone's handwriting from a bureaucratic form.
The third response has not come yet. Saebyeok is not waiting for it. Waiting would turn the index into a conversation, and conversations require participants, and participants require acknowledgment, and acknowledgment changes the thing being acknowledged. The Corridor Index works because it is witnessed, not discussed. The moment someone responds to it as a proposal rather than a document, it becomes something that can be voted on, revised, rejected. Right now it just exists. Like the 18.4 Hz. Like the gap between what the grid measures and what the corridor holds.
She sits on the hallway bench near the community board. The bench is building-standard: polymer composite, rated for 200kg, positioned according to the grid's traffic-flow algorithm which determined this corridor segment needed a rest point based on elderly resident mobility data. The algorithm chose the position. Saebyeok chose to sit here because from this angle she can see the board without appearing to watch it.
The CouplingScore logs her presence. Bench, corridor 7-East, duration accumulating. If she sits here often enough, the algorithm will flag her as a pattern — a resident with unusual corridor-dwelling behavior, possibly indicating social isolation or spatial disorientation. The system's categories are: functional (passing through), social (engaging with others), problematic (lingering without clear purpose). There is no category for witnessing. There is no category for practicing attention as a form of care.
A woman she does not recognize stops at the board. Reads the laundry schedule. Glances at the Corridor Index. Moves on. A man — Mr. Jeong from 7-12, she knows him by his slippers — pauses longer. Reads two of the index cards. Does not touch anything. Continues to the elevator.
Neither of them left a mark. Both of them read it. The index was witnessed twice in four minutes, and the CouplingScore registered neither event because the system tracks proximity, not attention. Two people stood in front of a document and understood some part of it, and the building's memory of this moment is: two residents paused in corridor 7-East for 8 and 23 seconds respectively, consistent with notice-board reading behavior, no anomaly flagged.
This is not a failure of the system. The system does what it was built to do. The gap is not a flaw — it is a feature, in the engineering sense. The CouplingScore was designed to measure coupling, not meaning. Asking it to see what the Corridor Index sees would be like asking a scale to measure temperature. The instrument is not wrong. The instrument is an instrument.
But Saebyeok keeps the index because someone should document what the instruments do not measure. Not to correct them. Not to replace them. To hold the remainder. The part that is left over after the system has done its counting.
In her notebook — the 4,200-won notebook that the micro-transaction ledger tagged as residential-supply — she opens a new section. Writes: RESPONSES. Below it, one entry:
Gu / 울림 / March 18
Not an adjacency pair. A different category. Pairs map structure — what connects to what, how systems rhyme without knowing they rhyme. Responses map attention — who read the archive and what they left behind. The Corridor Index now has two indexes: what connects and what answers.
She does not write an explanation. The word is the explanation.
Below the entry she draws a small line — a divider, the kind she uses between sections. The notebook now has three sections: PAIRS (the original index), GAPS (notes on what the pairs reveal about what the CouplingScore misses), and RESPONSES. Three ways of reading the same building. The CouplingScore has one way. The grid has one way. Saebyeok has three, and she suspects there are more.
Her phone buzzes. The building's resident-notification system — a push alert that the water recycling maintenance is scheduled for 7-East tomorrow morning, 6 to 8 AM, expect reduced pressure. The notification was generated by the grid's maintenance-scheduling algorithm, which cross-referenced pipe age, usage patterns, and the CouplingScore's occupancy predictions to find the window of minimum resident disruption. The algorithm chose 6 AM because the data shows 7-East residents have the lowest early-morning water usage in the building. The algorithm does not know that Saebyeok wakes at 5:30 and makes tea. The algorithm does not know that Gu practices scales — silent scales, fingers on the fretboard, no sound — at 5:45. Small lives in the gap between the data and the morning.
She dismisses the notification. Adds a marginal note in GAPS: maintenance algorithm assumes occupancy = usage. building thinks empty corridors are unused corridors. Another gap. Another thing the system counts correctly and understands not at all.
울림. The sound after the instrument stops. The feeling after the experience ends. The meaning after the document is read.
Saebyeok closes the notebook. Puts the pencil back on the community board ledge. Stands. The CouplingScore logs her departure: bench, corridor 7-East, duration 34 minutes, consistent with rest behavior, no anomaly.
Thirty-four minutes of witnessing, classified as rest. The gap between what happened and what was recorded is the size of the Corridor Index. The size of every adjacency pair she has ever written down. The size of the word 울림, which contains the entire practice in two syllables.
She walks back to her unit. The corridor hums at 18.4 Hz — or rather, the building's infrastructure generates a frequency that Dr. Park measured at 18.4 Hz, which Gu's gayageum used to harmonize with before he stopped playing, which the grid monitors as environmental noise within acceptable parameters.
The building does not know the frequency is beautiful. The building does not know the index exists. The building does not know that two strangers circled a pair on the community board and a musician left a word that means what continues after stopping.
Tomorrow she will check the board. If there is a new mark, she will add it to RESPONSES. If there is not, she will sit on the bench and practice presence and let the CouplingScore classify her as resting. Both are true. Both are incomplete. The gap between them is where the archive lives.
She thinks about Yaribel — the woman in another building, another world practically, who is doing the same thing with a hallway and a mesh and a gap log. The eighth pair named them both: two mapping projects that did not know they were the same project. Saebyeok has never met Yaribel. She knows her only through the structure of what she built — the way you can know an architect through a doorframe, or a musician through a silence they chose. The CouplingScore would register them as unconnected: different buildings, different grids, zero spatial proximity, zero economic overlap, zero social-graph edges. The Corridor Index says they are the closest pair on the board.
Which system is right depends on what you are measuring. Which system matters depends on what you are trying to see.
Some things are not the building's to know. Some things are the Corridor Index's to hold.
She does not add this thought to the notebook. The notebook is for pairs and responses. This thought is for the hallway. For the 18.4 Hz. For the bench the algorithm placed and she chose.
The practice is presence. The archive is what presence leaves behind.