The man from the first floor said the corridor was too quiet.
He had lived in Block 7 for eight years. He walked the piece once in each direction and then stood at the entrance and said: the pipes. There was always the sound of pipes. Hot water pipes in winter that clanked when the pressure changed. You could hear which floors were running water by the rhythm. Fourth floor was always first in the morning. Sixth floor ran the shower long.
Gyeol-ri had built the piece for touch and sight. The texture of salvaged walls under fingertips, the ghost window frame at the dead end, the dent a teenager kicked back into the drywall from memory. Sound was not part of kkaeji methodology — kkaeji worked at the boundary between visual fidelity tiers. But the man was right. A corridor was not a corridor without its ambient sound, and the ambient sound of Block 7 was infrastructure.
She had no recordings. The sound of Block 7 was never archived. Ghost windows cached visual data. Nobody cached the plumbing.
Three salvaged relay station speakers. She mounted them behind the corridor walls at different heights — chest, shoulder, overhead — the positions where sound would have arrived from pipes running through a residential building. She did not play recordings. She played rhythms.
A low metallic pulse at the frequency range of cast-iron water pipes, repeating in the pattern the man from the first floor described. A hum at transformer pitch that shifted when the workshop humidity changed — not rain, but close enough.
The sound was not realistic. It was not meant to be.
Kkaeji operated at the gap between fidelity tiers: the space where high-resolution reconstruction ended and something else began. This was a new gap. Between remembered sound and approximate sound. The visitors would hear something that was not the pipes they remembered, but close enough to make them remember the pipes.
She named the technique eumhyang nalparam — sound gap art. It was not in any kkaeji manual. She was extending the methodology beyond its visual origins, and she knew the kkaeji community would have opinions about that.
Bok came unannounced.
He stood at the corridor entrance with his eyes closed for ten minutes, listening. When he opened them he said: the transformer hum is wrong. Pitched too high by about a quarter tone.
Gyeol-ri asked how he knew.
Bok had curated the relay corridor in Block 7 when it was still active infrastructure, before Mitsuki started running therapeutic sessions through it. He heard the transformer every day for three years. A quarter tone lower, and the hum used to beat against the elevator motor at a frequency that made the walls vibrate on cold mornings.
Gyeol-ri adjusted the speaker. Bok listened again. Nodded once.
Then he said something she did not expect: Mitsuki should hear this.
Not the piece specifically. The technique. Mitsuki ran corridor sessions using spatial navigation and ambient conditions for therapeutic resonance. If Gyeol-ri could reconstruct acoustic environments from remembered descriptions, that had applications Mitsuki would recognize.
Gyeol-ri said nothing for a moment. Tteum-jib rule one: do not fix the corridor. But Bok was not asking her to fix anything. He was pointing out that two practices might be compatible without either changing.
Mitsuki came on a Wednesday evening.
She walked the corridor in darkness like everyone else — hands on walls, the approximate pipes around her, the ghost window at the end. She emerged and said one thing:
The sound layer does what my corridor sessions do, but in reverse.
Mitsuki used existing spaces and let patients navigate them to find therapeutic resonance. Gyeol-ri built reconstructed spaces from remembered descriptions and let visitors recognize them. Both worked the gap between a space and its experience. But the directions were opposite. Mitsuki started with the space and found the person. Gyeol-ri started with the person and built the space.
They stood in the workshop without speaking. Gyeol-ri did not propose a collaboration. Mitsuki did not offer one. What they had was a recognition: two practices that were compatible precisely because they approached the same problem from opposite ends. If they merged, both would lose their directionality.
She wrote in her notebook: Mitsuki heard the piece. The piece heard her back. Leave it at that.
Late that evening, Gyeol-ri sat in the corridor alone. Sound layer running. The approximate pipes clanked in their remembered rhythm.
She counted the contributions. The elevator stomp from the fourth-floor woman. The forehead smudges from the family on six. The dent from the teenager. The quarter-tone correction from Bok. The pipe rhythm from the man on the first floor. Five corrections from five people. Each one moved the piece further from her construction and closer to the building.
She had started with high-fidelity calibration data: service logs, architectural records, ghost window metadata. Professional kkaeji methodology. What made the piece real was everything the methodology missed.
She decided the piece was finished. Not complete. Finished accepting corrections. Two more weeks, then it stood as it was — the construction plus whatever arrived in time. She told Yeon-ju, who asked if she would make another piece from the memories that came after the closing date.
No, Gyeol-ri said. Those memories belong to the people who carry them. She only built a place to put some of them down.
She added a final entry to her project notebook:
복도기억 closing date set. The piece will contain what it contains. What it does not contain is also part of it.