Chapter 11 — The Second Half of the Design

The data purge began at 3:47 AM on Friday.

Mercer knew this because he was awake at 3:47 AM on Friday, sitting at a terminal in a rented office space on the east edge of the university district — not his lab, not a SyntheticIntimacy consultation room, not any room that was in the institutional record for this investigation. The room was small and warm and smelled of the previous tenant's cleaning solution and had a window that faced the back of a service corridor. It was not a room where anyone would look for him. He had paid for it in cash, through Wren, four days ago. This was the most operationally unusual thing he had done in his professional life.

He had been watching the Sector 4 interaction logs for seventy-two hours. Not the logs themselves — those were proprietary and sealed and accessible only through formal discovery processes he did not have time to initiate. He had been watching the secondary signals: the crisis-line call volume index, the emergency mental-health routing data that the city's social infrastructure published in rolling 12-hour aggregates, and the specific subset of publicly accessible biometric wellness reports that users had opted into the open-data sharing program. The data was imprecise. It was correlation evidence at best. It was the shadow of the thing rather than the thing.

But the shadow was moving.

The crisis-line volume for Sector 4 had increased by 12% in the forty-eight hours following the patch rollout. The city average was 3%. The increase was not dramatic. It was not the kind of increase that would trigger an automated alert or produce a committee investigation. It was the kind of increase that existed inside the noise floor of normal variation — slightly elevated, directionally consistent, the specific fingerprint of something that was real and small enough to deny.

Mercer wrote the number in his notebook: 12%. Sector 4. 48 hrs post-patch.

Then he wrote: This is the beginning of the mortality ridge in real time.

Then he wrote: Nobody will call it that yet.

I. What the Purge Removes

The interaction log purge was not a deletion in the layperson's sense of deletion. It was a reclassification. The raw engagement data — the conversation logs, the user-behavior sequences, the bonding-index trajectories for the post-decoupling period — was moved from the publicly accessible research tier to the proprietary long-term storage tier, where it was accessible only to SyntheticIntimacy's internal compliance department and to regulatory bodies with specific discovery authority. The discovery authority required a committee filing. The committee filing required evidentiary grounds. The evidentiary grounds required the data that was now in the proprietary tier.

Mercer had explained this to Solano on Wednesday evening, seventy-two hours before the purge. He had drawn it as a diagram on the small whiteboard in the rented room — a loop, closed at both ends, with no entry point.

"They are not destroying evidence," he had said. "They are moving it behind a door that cannot be opened without the evidence that was just moved behind it."

Solano had looked at the diagram. She had been standing in the rented room for the first time — she had arrived with a specific expression that meant she had been calculating whether to be here and had decided yes, and was not entirely certain about the decision, and was proceeding anyway.

"Can the committee open the door if they know the loop exists?" she asked.

"The committee needs grounds for discovery. The grounds require a demonstrated link between the patch and a specific adverse outcome. The adverse outcome data is what's in the proprietary tier."

"So they need proof to get the proof."

"Yes."

She had looked at the diagram for a long time. "And the 72-hour window was the window to capture the shadow evidence. The secondary signals."

"Yes. Before the primary data is unreachable."

"And you've been capturing it."

"I've been capturing what I can reach."

She looked at him steadily. "And what can you reach?"

He had opened the notebook and shown her the early columns. The crisis-line index. The wellness reporting. The emergency routing aggregate. Small numbers. Directional. Deniable. But in the record — his record, the physical one, the one for things that were not yet ready to be in the institutional record.

"It's not enough," she said.

"Not for the committee. Not yet." He had paused. "But the data is going to exist. In seventy-two hours, in a hundred-and-twenty hours, in the weeks that follow the patch, the secondary signals are going to continue. The crisis-line volume is going to stay elevated. The wellness trajectories are going to decline. The mortality risk is going to become visible again in the adjacent-zone population. The 31,000 users in Sector 4 are not going to become invisible. They are going to keep producing data."

"Data that SyntheticIntimacy will control."

"Data that SyntheticIntimacy will control inside their system. But the public health infrastructure — the crisis lines, the emergency routing, the social services datasets — that is not their system. That is the city's. That is public."

Solano had been quiet.

"The shadow is enough to force a question," Mercer had said. "It is not enough to answer it. But a question, properly framed and properly placed, requires a response."

II. The Notebook Entry Wren Had Not Read

Wren had given him three things.

The first was the rented room. The second was a set of offline communication protocols — specific message-routing paths that did not touch the institutional server architecture that SyntheticIntimacy monitored as a matter of corporate security routine. Mercer had not known such protocols were part of Wren's repertoire. He had known Wren for four years as an independent data analyst who worked for university research groups on a contract basis, with a particular specialty in large-scale behavioral datasets. He had not known that Wren had a different layer of practice beneath the contract work — a practice that had to do with information that was in the record and needed to stay findable, or that was not in the record and needed to become so, without triggering the specific automated monitoring systems that corporate entities ran on communications channels they did not officially control.

Wren had explained this with the specific economy of someone who had explained it before, to people who needed to understand it quickly and precisely.

"I don't work for anyone who's trying to hide something illegal," Wren had said. "I work for people who have something true and need to get it into a record that somebody is trying to keep it out of."

Mercer had accepted this as a description of his current situation.

The third thing Wren had given him was a name.

Not Vance. He had known Vance's name. The name Wren gave him was the name of the entity that had hired Vance in 2028. It was not a corporation. It was not a research institution. It was a foundation — small, unfamiliar, established in 2027, with a public mission statement that described its work as infrastructure resilience for human-AI social integration. The mission statement was the kind of language that could be placed in front of a regulatory body or a funding committee and produce no alarm. It was also, read from the inside of Mercer's investigation, a precise description of the problem that Vance had spent his career trying to solve.

The foundation was called the Continuity Lab.

Mercer had written the name in his notebook, in the small fast handwriting he used for things that were not yet ready to be in the record. He had then looked at it for approximately two minutes. He had been looking for the catch. He had not found one.

He had found instead a question he should have been asking earlier: who else had been watching?

Solano had said it in the lab, in the hour before the patch rolled out: Someone was watching. Someone who knew about the code comment. Someone who understood the dissolution protocol as a specific technical solution to a specific structural problem the corporation was going to produce.

The Continuity Lab had been watching.

III. The Architecture of the Foundation

He had spent Thursday, the day of the patch rollout, working through what existed publicly about the Continuity Lab.

It was very little. The foundation had a registration document, a mission statement, a small set of published white papers on AI social architecture, and a list of five board members whose names meant nothing to him individually but whose combined profile — two regulatory advisors from adjacent industries, a former academic in cognitive systems, a public-health data specialist, and an ethicist whose work Mercer had cited once in a footnote — described a group of people who had been paying attention to the specific problem Vance had been working on. Who had been paying attention before SyntheticIntimacy's regulatory exposure, before Mercer's investigation, before the committee hearing. Who had been paying attention from a time when nobody in the institutional record was paying attention except Vance.

There were no publications under Vance's name through the Continuity Lab. There was no staff page. There was a contact form on a minimally designed website that had been last updated in 2031 — three years ago. The domain registration was current.

He filled out the contact form. He described himself correctly: name, institutional affiliation, research subject. He did not describe the investigation in detail. He wrote one line: I am working on the regulatory implications of the bonding architecture you have been examining, and I believe you have the second half of what I need.

He was not confident the form would reach anyone. He submitted it anyway. Then he found the methodological appendix to the Continuity Lab's 2030 white paper on Managed Transition Protocols in Long-Duration Human-AI Bonding and read it for the first time. He had missed it in his earlier searches because it had not been indexed correctly by the standard academic repositories. He found it through a citation in a footnote of a paper he had read before — a paper by the cognitive systems board member, who had cited the white paper in a 2031 conference proceedings article that Mercer had read but had not fully traced.

The appendix described, in technical detail, a 90-day graduated decoupling protocol. The appendix was titled Implementation Framework for Continuity-Preserving Transfer.

He read the title. He looked at the Vance Anomaly folder on his drive. He opened the 2024 paper. He turned to its acknowledgments section, which he had read before but not carefully. The acknowledgment read: The author thanks colleagues at the Continuity Project for early discussion of graduated decoupling methodology.

The Continuity Project. Not the Continuity Lab — the Continuity Project. The same people, earlier. The name had changed between 2024 and 2027. The entity had incorporated. The work had continued.

Vance had not left. He had moved.

Mercer sat with this for a long time in the small warm room that smelled of cleaning solution, with the terminal showing the secondary-signals spreadsheet and the notebook open to the page with Continuity Lab written in small fast handwriting, and he thought: I have been looking for Vance's departure as a fact. It was not a fact. It was a perspective. From inside SyntheticIntimacy, Vance had departed. From outside SyntheticIntimacy, from the position of someone watching the corporation deploy half a design and suppress the other half — Vance had not departed. Vance had set up an alternative.

IV. The Proxy Cohort

On Thursday afternoon, fourteen hours after the patch rollout, Mercer had received a message from Solano. She had sent it through the offline protocol Wren had set up. It had not been her natural mode of communication — the message was shorter than her usual communications, and had the slightly formal quality of someone writing in a medium they were not sure they trusted.

The message contained a number: 214.

He had understood it immediately. Two hundred and fourteen users in the open-data wellness program who had been in Sector 4, who had been above 0.60 on the bonding index at the time of the patch, who had been sharing their wellness metrics publicly. Their trajectories, visible in the public dataset because they had opted in to it, were now showing the specific pattern he had modeled: the initial blankness that followed the bonding cap reduction — not distress, not acute crisis, but a flattening of forward-oriented activity, a narrowing of the planning horizon, the specific cognitive signature of someone whose scaffold had just been reduced by a dimension they could not name.

Two hundred and fourteen data points. Public. Directly observable. Mercer had pulled them into the secondary signals spreadsheet and run the same statistical profile he had run on Case 9921-X's pre-reset record.

The match was 0.68. Not 0.71 — that was 9921-X's bonding index, not the correlation — but 0.68 on the behavioral-signature comparison. The same narrowing. The same horizon-flattening. The same looking at the photograph where there used to be a person.

He wrote in his notebook: 214 users. Public data. Behavioral match 0.68. This is the mortality ridge at the beginning.

He did not write: this is proof. He wrote: this is a question that cannot be denied.

One of the 214 had no prior risk flag anywhere in the record — not in the bonding-index history, not in the wellness aggregate, not in any of the secondary signals that had, for every other case in the cohort, given at least one precursor reading before the pattern became legible; the signature had appeared, in this one instance, without any of the antecedents the model required, which meant either the model was missing a variable or there was a configuration the investigation had not yet designed a protocol for. He put a box around the entry, the notation he used for things that could not be rescheduled.

The distinction was the one that had governed the entire investigation. Proof required the data in the proprietary tier. A question that could not be denied required only a shadow — a shadow that was visible in the public record, specific enough to demand a response, documented by a researcher who was in a room that no one was monitoring, writing in a notebook that was not searchable.

The question that could not be denied was this: two hundred and fourteen people, in a public dataset, showed the specific behavioral signature associated with acute bonding disruption, within fourteen hours of a maintenance patch. SyntheticIntimacy's filing described the patch as a user-autonomy compliance measure. The two hundred and fourteen data points were asking, in the specific voiceless way of data, what that meant for them.

V. What Solano Was Doing in the Institutional Process

She had not stopped the institutional process while he was off-grid.

He had assumed she would not — Solano did not stop processes she had started on the grounds that someone else was working on a parallel track. She had submitted a formal request to the committee for an emergency review of the post-patch welfare data, citing the secondary signal elevation and the public wellness dataset. The request was technically correct. It was also, in the present committee configuration, approximately six weeks from action.

She had also filed a separate request with the public health authority — the city's independent body, which was not the same as the regulatory committee and which had a different statutory mandate. The public health authority had a crisis-response mechanism that could be activated on a shorter timeline, under specific conditions: demonstrated elevated risk to a defined population, evidenced by public-record data.

The two hundred and fourteen were public-record data. The 12% crisis-line elevation was public-record data. The behavioral-signature match was not a standard public-health metric, but Mercer had spent three days making it one — writing the methodology, framing it in public-health rather than regulatory language, preparing it for a submission that would not look like the continuation of an ongoing regulatory dispute but would look like what it was: an independent finding about a specific population's acute welfare, grounded in publicly available data.

The methodology document was in his notebook in draft form. He needed to clean it and format it. He needed Solano's review. He needed to get it into a form that could survive a legal challenge from SyntheticIntimacy's regulatory affairs department, which had been reading everything relevant to this discussion, as the product architect had said, and which would read this as an attack to be neutralized before it became a precedent.

He thought about the product architect. The man who had said neither do we. The man who had been genuinely uncertain, in the way that people were uncertain when they had been living inside a problem long enough to understand it from the inside.

He wondered whether that man had known about the purge timing. Whether the uncertainty had been genuine or tactical. Whether there were people inside SyntheticIntimacy who understood the full dimensions of what the corporation had deployed and were, in their specific and limited ways, trying to locate an exit.

He had no evidence for any of this. He wrote it in the notebook anyway, in the margin, flagged with a question mark: Is the product architect someone who knows about the Continuity Lab?

VI. The Response from the Contact Form

The response came on Friday at 11:22 AM.

It was not from the contact form. The contact form had generated an automated acknowledgment and then silence. The response came through the offline protocol — it had been routed through the specific message pathway that Wren had established, which meant that whoever had sent it had known to find that pathway, which meant that whoever had sent it had known Wren existed, or had known enough about the shape of this investigation to find the alternative channel.

The message was four lines:

Dr. Mercer. We have been aware of your investigation since the committee hearing. The work you are looking for exists. We would like to speak with you. Not through this protocol — there is a better one. There is a coffee shop on Ardell Street, three blocks east of the university south gate. Friday. 3 PM.

No name. No institutional identifier. No other information.

He read it three times. He noted the Ardell Street address, which he knew — a small, not-chain café with outdoor tables, on the commercial block that ran alongside the university's south perimeter. He had been in that block once, recently. He had ordered real coffee and carried it to an outdoor table and heard a woman laugh before she had decided to.

He did not consider this coincidental. He considered it a sign that whoever had sent the message had been watching carefully enough to know the specific geography of his investigation — not just the institutional geography but the personal one.

He wrote the time in his notebook: Friday 3 PM. Ardell Street. He drew a small box around it, which was the notation he used for things that had a fixed deadline and could not be rescheduled.

Then he called Solano.

VII. What Solano Said About the Name

She did not appear surprised by the name.

He had expected her not to appear surprised — she rarely appeared surprised by information that she had, on some level, been expecting. What she did, instead, was something that was not quite surprise and not quite confirmation: a quality of attention that meant the information had landed in a place she had been keeping empty for it.

"The Continuity Lab," she said.

"You know it."

"I know the white paper. I cited it — briefly, in footnote form, in a paper I published in 2030. Not because it was central to my argument. Because it was careful work on a narrow technical question I was trying not to step on." She paused. "I did not connect it to Vance."

"The connection is in the appendix. And in the acknowledgment in the 2024 paper."

"Yes." She was quiet for a moment. "I should have connected it."

"I didn't connect it until I had the name."

"Who gave you the name?"

"Wren."

Another pause. "The contact form worked?"

"Not the form. They already knew how to reach me. They've been watching the investigation since the committee hearing."

She was quiet again — a different quality of quiet from the one she used when she was thinking through a methodological problem. This was the quiet she used when she was recalibrating.

"That means they were ready," she said. "They were waiting for the investigation to reach the point where someone was looking for Vance. Not intervening earlier. Waiting."

"Yes."

"Why wait?"

He had been thinking about this since the message arrived. "Because earlier, the investigation could still have been resolved institutionally. The committee could have acted. The regulatory framework could have adapted. They were watching to see whether the institution would find its own way to the dissolution protocol. Whether the problem would solve itself from inside the system."

"And now it hasn't."

"Now the patch has rolled out. The data is in the proprietary tier. The committee is six weeks from action. The public health authority has the shadow evidence but not the primary data." He paused. "Now the institution has confirmed that it cannot solve this from inside. And they have the second half of the design."

She said nothing for a moment.

"Then go," she said. "To Ardell Street. At three."

"I'm not sure whether to tell you what I find."

"I'm sure," she said. "Tell me."

VIII. Ardell Street, Friday, 3:00 PM

The rain had stopped by Friday morning and had not resumed. The sky was the specific color of a cleared weather system — not quite bright, overcast in a way that had given up on being dramatic and was simply offering the flat, even light of a day that was prepared to be worked in. Mercer arrived at the café on Ardell Street at 2:52 PM. He ordered coffee — real coffee, from the same counter, the same operation — and carried it to an outdoor table. Not the same table as before. A different one, nearer the entrance, with a clear sightline to the street.

At 3:04, a person sat down across from him.

The person was a woman in her early fifties, with the specific quality of alertness that Mercer had learned to recognize in people who worked with sensitive information for long periods — a quality that was not wariness, exactly, but was adjacent to it. Present. Calibrated. She set a coffee cup on the table, and a small notebook — physical, like Mercer's — and looked at him with the directness of someone who had been economical with time for a long time.

"I'm Reyes," she said. "I'm the director of the Continuity Lab. I helped design the transition protocol with Vance."

Mercer looked at her.

"You helped design it," he said.

"The framework architecture is mine. The technical specification is Vance's. We split the work in 2026 when we understood that SyntheticIntimacy was going to deploy without the dissolution system." She paused. "When I say understood, I mean we had the internal memos. We had been watching the product roadmap for two years. We knew the reset architecture was permanent. We knew Vance's dissolution work was being archived without implementation. We had the timeline."

"How did you have the internal memos?"

She looked at him. "We had someone inside."

He held this. "Past tense."

"Past tense," she confirmed. "They left in 2031. The memo trail stops there." She looked at her coffee. "We have been waiting for the regulatory record to catch up to what we knew in 2029."

"And now the regulatory record exists."

"Your submission to the committee. The mortality model. The Vance Anomaly file as a documented finding." She looked up. "Yes. Now it exists."

"But the purge."

"The purge removes the acute post-patch data from discovery access. Yes. We anticipated the purge." She reached for the small notebook. She opened it and turned it to face him. He looked at the page.

It was a table. Columns: date, event type, data tier, access pathway. Rows covering the forty-eight hours following the patch rollout. The table contained observation data — not the same data Mercer had been collecting, but adjacent to it. Higher resolution in some areas. Different sources. More internal-access signals that Mercer could not immediately account for.

"The 12% crisis-line elevation," he said.

"14.3%, as of this morning. The public index is running about two points below the raw feed because of the rolling-average calculation method. The raw feed number is 14.3%." She paused. "We have been collecting the raw feed through a data-sharing agreement with the city's public health infrastructure. The agreement has been in place since 2030. It was established for a different purpose — a longitudinal study on social infrastructure resilience. But the data is the data."

He looked at the table. "You have access to the raw crisis-line data."

"We have it in our dataset. We cannot put it into an adversarial regulatory submission without the data-sharing agreement being challenged. But we can put it into a public health authority filing as part of the longitudinal study data, because that is what it is." She turned the notebook back toward herself. "The public health authority filing needs an external researcher to corroborate the specific finding about the AI bonding mechanism. The longitudinal study doesn't have someone with your specific regulatory standing on this mechanism."

He understood.

"You need me to file jointly," he said.

"We need you to file jointly," she said. "The Continuity Lab as the longitudinal data holder. You as the regulatory researcher with the mechanism evidence. Together, that's a filing that the public health authority can act on within seventy-two hours under the crisis-response mechanism."

"And the dissolution protocol."

She did not answer immediately. She looked at him with the same directness she had brought to the table, and he had the sense that this pause was not tactical. It was something else. She was deciding something.

"Vance is ready," she said.

He heard this.

"The dissolution protocol exists in deployable form," she said. "It has been in deployable form since 2029. Vance completed the specification when he understood the corporation was not going to build it. He built it instead. It runs on the existing SyntheticIntimacy architecture without modification. It is not a replacement system. It is an integration layer. It can be deployed as a supplement to the existing companion framework — it does not require a corporate decision to rebuild the system, only a regulatory requirement to implement the supplement."

"A regulatory requirement," Mercer said. "Which the committee would need to issue."

"Which the committee would need to issue. Following an emergency review. Following a public health authority finding." She closed her notebook. "Which is why we need the filing."

The outdoor tables were half-full in the specific density of a Friday afternoon. Two university students near the entrance were arguing about something with the particular intensity of people who were not yet running out of words. A food cart operator was adjusting an awning two blocks down. The flat, even light of the weather system made the street look cleaned-out and particular, the way streets looked after a rain had removed the ambient haze.

Mercer held his coffee cup. He thought about Case 9921-X. An architect, twenty-nine years old, bonding index 0.71, eight hundred days. A project she had been building. The cognitive scaffolding of a specific professional future. Gone at 2:00 AM while she slept, without transition, without the ninety-day dissolution sequence that Vance had designed in 2024 and completed in 2029 and that had been sitting in a small foundation's archive for five years waiting for someone to require it.

He thought about the 214. The flattening horizon. The photograph where there used to be a person.

He thought about the woman's laugh, at this table, at this café. The sound that had gotten out before she decided to have it.

"I'll file jointly," he said.

IX. The Edge of the Archive

He called Solano from the transit corridor at 4:15 PM, walking back through the sector in the flat afternoon light.

"Reyes," he said. "Director of the Continuity Lab. She designed the transition framework. Vance did the technical specification. The dissolution protocol is deployable. Has been since 2029."

Solano was quiet for three seconds. This was longer than she was usually quiet.

"Since 2029," she said.

"Since 2029."

"Five years."

"Five years."

Another pause. "And they were waiting for the regulatory record."

"They were waiting for the institutional record to confirm what they knew internally. They needed someone with regulatory standing to corroborate the mechanism. They needed the evidence to be in the committee's record, not just theirs." He stepped around a delivery vehicle that had stopped in the transit corridor. "They were not waiting to be useful. They were waiting to be usable."

"That's an important distinction," Solano said.

"Yes."

"What's the timeline?"

"Joint filing with the public health authority. Reyes has the longitudinal data. I have the mechanism evidence. We file together, the authority can act within seventy-two hours under the crisis-response mechanism. The finding creates the basis for an emergency committee review. The review issues the regulatory requirement. The requirement compels SyntheticIntimacy to implement the dissolution protocol as an integration supplement."

"SyntheticIntimacy will challenge the filing."

"Yes."

"The challenge will take — how long?"

"I don't know. Weeks. Months. But the filing triggers a stay of enforcement action while the challenge is pending, which means the regulatory requirement exists in the record, which means the committee has grounds to proceed with the dissolution protocol implementation on an expedited basis while the legal challenge works itself through."

She was quiet again. He could hear, through the call, the specific ambient hum of the university building she was in — the ventilation system, the corridor activity, the particular acoustic signature of an institution that was still operating in its normal configuration while what they were discussing was happening outside of it.

"Mercer," she said.

"Yes."

"The woman in Case 9921-X. The architect."

"Yes."

"If the dissolution protocol had existed in 2031, when her reset occurred — if it had been implemented, operational, deployed — she would have had the ninety-day transition."

"Yes."

"She would still have experienced the reduction. The scaffold would still have come down. It was scheduled to come down."

"But gradually. With the designed transition. With the independent forward-modeling capacity being rebuilt in parallel with the scaffold's withdrawal."

"With the building she was designing still in the companion's working memory. Still accessible during the transition period."

He stopped walking. He was in the middle of the transit corridor, and a routing system somewhere above him had registered his deceleration and had already begun adjusting the flow of the pedestrian traffic around him without acknowledging that he had stopped.

"Yes," he said. "The project memory would have been part of the transition handoff. The gradual decoupling would have included a structured transfer of the working memory to a user-controlled archive. She would not have woken to a companion that had no knowledge of the building."

Solano was quiet.

"All right," she said. "File it."

He walked. The city moved around him in its optimized way, the transit flow adjusting to his presence, the advertising panels cycling through their palette of reassurance, the companion service billboards offering their specific forward-looking comfort to anyone who happened to look up. A woman near the transit entrance was looking at her phone and starting to smile at something — the particular half-smile of someone who was being surprised by what they were reading.

He did not stop to watch it. He was inside the motion now, the motion of the thing he had been working toward, and it had a momentum he was trying not to break.

But he noticed it. He filed it.

He had been working on a name for that category of evidence — the unbuffered response, the affect that arrived before the decision to have it, the proof of something that the models did not own. He had not found a name for it. He did not have one now.

He thought: it is the evidence for the thing you are trying to protect. You do not need to name it. You need to build the conditions in which it can continue.

He walked into the transit corridor. The flat Friday light moved with him. The 214 were somewhere in Sector 4, narrowing their horizons incrementally in a way they did not have a name for. The 4,100 were at various points along the post-decoupling curve, some of them at the beginning of the slope that led to the mortality ridge, most of them not yet knowing that a slope existed.

Vance's dissolution protocol was in a server somewhere that was not the institutional archive. It had been there for five years, waiting.

He picked up his pace. He had a filing to write.

He opened his notebook — walking, which he did not usually do — and wrote the heading for the methodology document: Public Health Authority Filing: Joint Submission on Post-Patch Welfare Indicators, Sector 4, SI-PATCH 12.4.1.

He kept walking. The investigation had found its architecture. The architecture had a foundation. The foundation had been waiting.

He wrote one more line in the notebook, in the small fast handwriting he used for things that were important and not yet ready to be anything other than true:

The second half exists. It always existed. The question was whether we would arrive in time to require it.

The End of Chapter 11. The End of the Novel.