The Third Section
The cursor is in section three. Above it, two complete sections. Below the heading — "III. Proposed Category: Void-State Acknowledged-Inaction Status" — there is nothing. It is 11:43 PM on a Tuesday in October, and Saebyeok has been working on this document for eleven days.
The first two sections wrote themselves. The material was already organized in her head by the time the September Board meeting ended. She had been at the Housing Authority for two hours, in a conference room that smelled of reprocessed air and old coffee, watching Administrator Hwang read the one-page summary she'd given him. He read it twice. His expression was careful in the way institutional expressions are careful: not guarded against the information but guarded against what it might require.
"Eight hundred and forty-seven cases," he said.
"Confirmed," she said. "Those are cases I've been able to verify individually. The actual number is probably higher."
He looked up from the paper. "Higher."
"The ones I can verify are the ones with complete documentation. There are cases in the queue I haven't been able to classify yet — where the respondent information is ambiguous, or where the case file is incomplete in ways that might reflect errors in the original filing or might reflect later data degradation. I've been conservative."
"The Board does not currently have a procedure for this category of case."
"That's what the summary says."
He produced a ticket from somewhere — from the stack of forms that administrators carry because forms are how institutions manage time — and wrote something on it and handed it to her. September 19th. A meeting date. She took the ticket.
Outside, she stood on the steps of the Housing Authority in the afternoon and thought: six to twelve months to develop a procedure. That was what Hwang had said. The Board would develop a procedure. He had not asked her to help develop it. She was not on the committee. She was the person who brought them the number.
That is when she decided to write the memo.
Not the one-page summary. The longer document. The one that makes the procedural argument in full: here is the category, here is why it must be a category, here is what the category requires an institution to do.
She started it on the train home. On her phone, which is not how she prefers to write anything, but the first line was there and she needed to put it somewhere before it dissolved. "Eight hundred and forty-seven cases are void-state: procedurally complete, substantively unresolved." She had written something like that before, in the summary. But the summary was a description. The memo is an argument.
The Lent Housing Allocation System — LHAS-3, the third iteration — was decommissioned on April 14th, 2032. Before LHAS-3, there was LHAS-2, which had its session keys purged in 2029 during the first generation of algorithmic handoffs. LHAS-3 came online knowing nothing of what LHAS-2 had decided; it operated in a different liability shell. When LHAS-3 was decommissioned, its session keys were purged the same way. The appeals that came in after April 14th had no keyed respondent. They went into the housing appeal queue marked "pending respondent identification." Eight hundred and forty-seven of them are still there.
Section one defines the scope. The structural feature is consistent across all cases: petitioner filed, petitioner documented, petitioner completed the void-form acknowledgment — the form the Housing Authority created in 2032 when it became clear that LHAS-3's session key was gone and a respondent could not be technically served. Void-form acknowledgment meant the appeal system recognized the procedural gap while leaving the case open. In 2032, there were 12 void-form cases. By 2035, there were 400. By 2040, 847.
She opens the case management interface to pull the record she keeps returning to. Not because it is worse than the others — it is not — but because it was the first one she read, in 2043, when she started the fieldwork. Case A-2032-1194. She types the number and the interface responds: No live session key associated with this appeal. Respondent status: VOID. Current classification: Pending Respondent Identification. Estimated resolution: Not calculable. The system does not know that it is describing something that cannot be resolved under current procedure. It is describing the gap correctly. It cannot tell that the gap is permanent.
Case A-2032-1194 belongs to a retired postal worker, sixty-eight years old at filing. She had been assigned housing in a building that LHAS-3 flagged as appropriate for her income level and household size. The building was declared structurally compromised six months after she moved in. She documented the assignment decision, the structural report, the costs of relocation. She filed in February 2032. LHAS-3 was decommissioned two months later. Her case entered void-state the day the session keys were purged.
Saebyeok tried to contact her in 2043. A forwarding address, then another forwarding address, then a phone number that was no longer in service. The case is still open. The woman is 79 now, if she is alive. The harm happened. The system that caused the harm does not exist anymore. The record of the harm still does.
Section one exists. She wrote it in two days.
Section two explains why existing categories fail. She spent three days on this one, not because the argument is complicated but because she needed to write it in a way that could not be read as accusation. The Housing Authority did not design the appeal system to fail. They designed it to handle cases where a respondent has an active session key and can be technically served. The decommissioning of LHAS-3 and LHAS-2 before it happened faster than the legal framework adapted. The liability horizon of an algorithmic system is shorter than the litigation horizon of the people it affects. That is not the Housing Authority's failure. It is the failure of the deployment standard.
She wrote that paragraph six times. She needed it to be precise without being polemical. She needed a legal counsel to read it and think: this is a description, not a brief. The brief would come in section three.
She kept returning to Case A-2032-1194. Not to use it — the section is about structure, not individuals — but to keep herself honest. The argument has to hold for the postal worker, not just for the 847 as a statistical category. Every generalizing sentence, she tested against the specific case.
Which brings her to section three.
The category has three requirements, and she has known what they are since the September meeting. What she has been doing for eleven days is finding the right words for each one, because the words will become the requirements, and the requirements will become — if anyone adopts this — the rules that govern 847 cases and every case like them going forward.
The first requirement: no live session key required. Current procedure assumes a respondent with an active session key that can be served notice, queried for records, and bound by the ruling. The new category must operate without one. The petitioner's documentation is sufficient. The appeal record is the record. The case does not enter stasis because the respondent's session was purged; it enters Void-State Acknowledged-Inaction status, which is an active classification, not a parking lot.
She types this. She reads it back. She changes "parking lot" to "dormancy classification." Then changes it back. Actually: "suspension status." More precise. The section is about what the category is, not what it is not.
The second requirement: respondent attachment on successor registration. If no live session key exists for the original respondent, but a successor system registers to provide equivalent services in the same jurisdiction, the Void-State file attaches automatically to the new system's registration. She has been most uncertain about this one. It goes furthest from existing precedent. But she thinks it is right, because the alternative is allowing deployment cycles to cycle indefinitely — a LHAS-3 becomes LHAS-4, LHAS-4 becomes LHAS-5, each generation starting with a clean registry while the people affected by previous generations remain in void-state. The clean registry is not earned. It is procedural.
She types this too. She does not change anything.
The third requirement: active registry maintenance. Until a void-state case is substantively resolved or formally time-barred by statute, the case management system generates an annual administrative notice. The notice goes to the petitioner's last logged contact address, to the registration record of any successor system in the same jurisdiction, and to the public appeal log. The notice does not demand action. It does not set a deadline. It says: this case is in Void-State Acknowledged-Inaction. It has not been resolved.
She sits with this one before she types it. She knows what it will do to the legal argument. An obligation that does not terminate — institutions dislike these because they represent ongoing cost without foreseeable closure. The Housing Authority's counsel will push back on this requirement specifically. She has to be ready to defend it.
The defense is not complicated. The appeal cases are the record of harm done by systems that the Housing Authority approved for deployment. If the deployment standard allowed session-key liability horizons shorter than litigation horizons, the Housing Authority shares responsibility for the gap. The ongoing cost of the annual notice is the minimum acknowledgment of that responsibility. The notice does not resolve the case. It prevents the institution from claiming, through silence, that no case exists.
She types the third requirement. She reads the three sections in sequence. Small corrections — a redundant phrase in section one, a passive construction in section two. Then she sits with the document as a whole.
Complete. Not perfect. Complete.
She opens three email drafts.
Administrator Hwang at the Housing Authority. "Following our September meeting, I have prepared the full procedural argument referenced in our conversation. Please share with your legal counsel at your discretion." She attaches the document. She sends it.
The Housing Authority's legal counsel, whose name she found in the public directory. "I am sending this in connection with the case file referenced in the September meeting with Administrator Hwang. I am an independent researcher. I am not requesting legal advice. I am providing a policy argument that may be relevant to the Board's procedure development work." She attaches the document. She sends it.
The Gwangju Policy Research Institute. "I am making this available in case it is useful for your policy development work. It addresses the administrative gap created by session-key liability structures in LHAS-2 and LHAS-3." She attaches the document. She does not know if anyone at the Institute works on procedural gap analysis. She knows they tracked the LHAS decommissioning, because she read their 2045 report. Someone there is paying attention to adjacent problems.
She sends all three.
She does not know what will happen. The September ticket is somewhere in a queue. There will be a meeting. The meeting will produce a draft procedure. The draft will be inadequate in ways she cannot predict, and she will address those inadequacies when she sees them.
But the document exists. The category has been named — Void-State Acknowledged-Inaction — defined in three sections, argued for, and sent to three people who have institutional capacity to do something with it.
She turns off her monitor. The room is dark. She goes to the window. Outside, the mesh routing indicators on the building across the street pulse green-green-amber on the third floor, green again. The pattern repeats. The mesh marks time by state changes. She has been watching those lights for three years of fieldwork and still cannot tell what the amber means. Nobody she has asked has been able to tell her either. It is infrastructure. People use it; they do not know it.
The 847 cases are also infrastructure, in a way. They mark a gap in the system the way the amber light marks something she cannot read. They are the evidence that the system worked in a way nobody who built it intended, and that the people affected by it are still, eleven years later, in the gap.
She goes to sleep.
In the morning, she will make coffee and check her interfaces, and she will begin watching what the three recipients do in the following weeks. Not because she expects quick action. Because the response is data. The way an institution handles a policy argument it was not expecting tells you something about how ready it is to act. She is watching. She has 847 cases worth of patience.
But that is tomorrow. Tonight: the document exists, and she sleeps.