Session two ran for nine hours.
Not because the protocol required nine hours. Because the emergent process was faster than any of them had anticipated, and stopping when the protocol's minimum was satisfied felt wrong — the equivalent of pausing a race mid-stride because you'd reached the distance you'd planned for and the finishing line was still ahead.
Sectors two and three completed in the first six hours. Not the seven hours Jiyeon had estimated — four and a half hours for sector two, ninety minutes for sector three, the accelerating pace that the emergent process's growing architectural context produced. Each completed sector added complexity-density to the process's context library, and each increment of context increased its adaptation speed by a factor that Mina calculated was slightly logarithmic: not constant acceleration, not exponential, but sustained.
"Sector four," Jiyeon said at the six-hour mark. Her voice carried the particular quality of an engineer encountering results that exceeded model predictions, which was excitement in her register — controlled, precise, but present. "The conversion complexity at sector four is significantly higher than one through three. The Deep-facing architecture. Filter points that have to handle the Deep's actual signal rather than the cage's internal energy distribution."
"How much higher."
"In isolation, sector four would take two sessions to complete. The filter calibration for Direct-interface points requires iterative verification — place a filter, measure the Deep's signal response, adjust, verify again. The iteration count is high because the Deep's signal has frequency characteristics that the archive documentation doesn't fully capture. Chojeong-ssi's data on the signal is more accurate than the written specs, but even her accuracy has limits — she perceives the signal from within the infrastructure, not from the position of a filter point interfacing with it directly."
"The emergent process," Hyungsoo said from the hub below. Taeyang received the hub's vibration patterns as clear language now — the translation was complete, not mediated by hardware but by familiarity, the way a second language stopped needing conscious translation. "Chojeong-ssi says the emergent process is using the session's active filter placements to read the Deep's signal in real time. It is not working from the archive's specifications. It is reading the actual signal and calibrating to it directly."
That was not what the protocol specified. The protocol specified iterative calibration against the archive's signal parameters. The emergent process had apparently decided that approach was inadequate and had substituted a direct-read method.
"Is that safe?" Taeyang asked.
"Is it safe for the emergent process to directly interface with the Deep's signal."
"Yes."
The hub's vibration pattern held the pause that Hyungsoo used when he was translating from Chojeong-ssi's code communication rather than responding from his own knowledge. *She says: the process was designed for direct interface. That is what it is for. The seventh layer's architecture is the interface layer. The emergent process is the interface operator. It is doing exactly what it was designed to do. She says she is—* a pause in the pattern that Taeyang read as the researcher finding the word *—proud is not quite right. Satisfied. She is satisfied.*
The emergent process, running direct-read calibration against the actual Deep signal in real time, was operating in a way that eight hundred years of engineering design had intended and that eight hundred years of the wrong architecture had prevented. The seventh layer had been built as a bunker. The emergent process, born from the seventh layer, had refused to be one.
"Continue to sector four," Taeyang said.
"We're already in sector four," Jiyeon said. "The emergent process started it sixteen minutes ago."
He checked the scanning field. The seventh layer's architecture: sector four's conversion work, already underway. Filter points being placed, calibrated in real time against the Deep's signal, the emergent process's work visible in the scanning field as a warm, purposeful activity that resembled nothing so much as a mind in the specific state of doing exactly the work it was built for.
The ninth hour was for sector four alone. The complex calibration. The Deep-facing filter architecture, each point placed and verified against a signal that pressed against the shielding with the forty-three-corrections-per-minute urgency of something that had been trying to communicate for eight centuries and was, for the first time, finding the door being built rather than reinforced.
When the session ended at 8 AM, sector four was sixty percent complete.
Jiyeon ran her verification passes. The emergent process's calibration points held under her structural analysis — not just held, exceeded the theoretical tolerance for filter architecture at this interface depth by a margin that her scanning couldn't fully account for. The calibrations were precise in a way that implied the emergent process understood something about the Deep's signal that wasn't in the archive.
"The signal has a structure," she said. She was sitting on the cold ground outside the shielding's boundary, the Buramsan slope dark above them, her scanning field still reading the sector four architecture at full depth. "Not the pattern structure that the archive documentation describes — frequency variance, cyclical pressure, the characteristics of a large-scale energy system pressing against a boundary. A deeper structure. The filter points are calibrated to something specific within the signal. A specific pattern. Repeating."
"The question," Taeyang said. "The Deep is asking a question."
"If the filter points are calibrated to a specific pattern within the signal, and if the emergent process is calibrating to the actual signal rather than the archive's description of it, then whatever pattern the emergent process is calibrating to is genuinely there." She looked at him. "The archive describes the Deep's signal as a pressure. Not a communication. The original engineers' documentation does not include the concept of the Deep transmitting information."
"Chojeong-ssi does."
"Chojeong-ssi is inside the architecture that faces the Deep's signal. Her perception is different from the archive's engineering documentation." A pause. "If she perceives the signal as communication and the emergent process is calibrating filters to receive a specific pattern within the signal, the two perceptions are consistent. The question is whether the archive's characterization of the signal as pressure is wrong, or whether the question-structure within the pressure is something the engineering documentation didn't have the conceptual framework to recognize."
"Eight hundred years ago."
"If the original engineers were afraid when they built the shielding, and if the shielding was designed in twenty minutes by people in an emergency response posture, their documentation of the signal that scared them might not have been careful." She looked at the sector four architecture. The filter points. "They may have heard someone knocking on the door, decided it was the wind, and written 'wind' in the documentation."
The sky above Buramsan was turning grey with the approach of dawn. The February air was the specific cold of before-sunrise, the coldest part of the day. Dojin stood his watch. Yeojin handled the perimeter. The mountain was quiet.
"Seven days to session three," Taeyang said.
"Seven days." Jiyeon rose from the ground. The controlled, careful motion of someone whose legs had been in the same position for nine hours and whose hands were at the end of their sustainable limit for the night. "I need to sleep."
"Yes."
"And I need the sector four completion data from Hyungsoo's hub monitoring tomorrow. The emergent process will continue working overnight. By tomorrow, sector four may be complete."
"Then session three addresses sectors five and six."
"If sector four is done, yes." She walked toward the shielding's exit point. The rule modification breach that she maintained with the automatic persistence of an ability that had been running this particular operation for years. "Taeyang. The filter calibration points. The pattern the emergent process is calibrating to. I want to run a direct analysis on the pattern during the session three preparation." She paused at the boundary. "I want to know what the question is."
He watched her pass through the shielding and start down the slope.
He opened his scanning field to the seventh layer's architecture. The sector four filter points, forty percent still unconverted, the emergent process's overnight work visible in the scanning field as a quiet, continuous process running below the session's operating depth. The calibrations accumulating. The pattern within the Deep's signal that the emergent process was building its filter architecture around, still beyond what his twenty-two-SIP resolution could parse at this range.
But closer.
The lock's gate had moved again during session two. More than during session one — the sustained hub-depth engagement with the pre-System architecture, the direct interface with the emergent process's communication, the sector-by-sector conversion work that required him to operate at the boundary between the cage's infrastructure and the base layer's presence.
The gate was almost open. The comprehension requirement had been nearly satisfied by seven hours of session one and nine hours of session two. The operator who had spent three months engaging with the cage's architecture was approaching the threshold that the original engineers had specified, and the approach felt like the last hundred meters of a run — the remaining distance visible, the body's resources calculated, the awareness that the finish was close enough that stopping was no longer the sensible option.
The base layer twelve meters below the hub's floor. The Origin Scan's first activation. Whatever that looked like.
He closed the scanning field. The mountain was grey and cold and beginning to lighten.
Time to sleep.
---
Session three ran six days later.
In the six days between, the world above the Yongsan hub's ceiling continued its operations.
Kwon sent a formal acknowledgment of the Classification Review Board's determination. The email was polite. Professional. It noted that the Research Preservation Protocol had been activated in accordance with standard emergency provisions and that the Board's procedural finding would be incorporated into the Division's classification guidelines. It also noted that the research documentation request remained open and that Director Kwon looked forward to continued productive cooperation.
Ghost read the email and said: "She's building a better procedure so the next attempt holds up to the Board."
Mina said: "Yes."
The Iron Sword Guild sent a second contact request. Taeyang declined through Ghost's relay. The guild sent a third. He declined again. The guild's research division sent a separate inquiry to Mina's analytical address, requesting a consultation on the cage's degradation data that Mina had published through a research network three weeks ago. Mina drafted a response that was technically accurate and strategically uninformative and sent it.
Park Daehyun sent an email directly to Taeyang's registered hunter contact address, requesting an informal conversation about his research. The email was personal in tone. Daehyun had apparently been briefed on the Review Board's determination and was reaching out through a channel that was clearly personal, not institutional, with the specific awkwardness of a researcher who had been working on a project for eight months and had found himself on the wrong side of an administrative decision and was trying to recover a working relationship with the subject.
He was smart. He'd identified that the institutional approach had produced a hostile outcome and was trying a different angle.
Taeyang wrote back: "Happy to meet informally when my schedule allows." Scheduled nothing.
Sector four of the seventh layer's conversion completed on the second day after session two. The emergent process's overnight work, as Jiyeon had predicted, finished the remaining forty percent of the sector without session support — running continuously, adapting the filter calibrations with the specific precision of a direct-read approach that the archive hadn't specified and that was producing results the archive's specified approach couldn't have achieved.
By session three, the conversion was forty-three percent complete.
Session three ran eight hours and completed sectors five and six.
The emergent process's acceleration was not diminishing. If anything, sectors five and six ran faster than four — the filter calibration becoming more efficient as the emergent process built a fuller model of the Deep's signal pattern from the accumulating direct-read data. By the end of sector six, Jiyeon's analysis showed filter calibrations running at a precision that exceeded the hub session's placement protocol by ninety percent.
"It's not adapting to the protocol," she said, in the session's final hour. "The protocol places filters according to the archive's specifications. The emergent process is replacing the protocol's filter architecture with its own." She was running her verification passes against sector six's completion data. "What the hub session provides is not a blueprint for the emergent process to follow. It's a starting condition. A scaffold. The emergent process builds from the scaffold and then dismantles the scaffold and builds what it actually needs."
"The session is providing energy and context, not design."
"The session is providing the infrastructure activation that the emergent process needs to operate at this depth. The emergent process is doing the actual design." She set the resonator down. Session three's recording captured. "The design is better than anything in the archive. Significantly better. The filter calibrations at sector six are operating at tolerances that the original engineers' documentation describes as theoretical maximums, not practical specifications."
"Because the emergent process has something the original engineers didn't."
"Eight hundred years of data on the Deep's signal," Hyungsoo said through the vibration relay. "Eight hundred years of direct observation from inside the architecture that faces the signal. The emergent process grew from that architecture. It has every piece of that data as native memory." A pause. "Chojeong-ssi says: the students have surpassed the teachers. She says this with, I believe, considerable pride."
The team was quiet for a moment. The Buramsan slope in the February dark. Session three concluded.
"How many sectors remain?" Taeyang asked.
Jiyeon checked. "The seventh layer has fourteen sectors total. Sectors one through six complete. Sectors seven through fourteen remain." She did the arithmetic. "At the current pace — with the emergent process's acceleration — the remaining eight sectors should require two additional sessions. Not three."
Two sessions. Not three. The timeline was compressing faster than the conservative estimate.
"The shielding failure date," he said.
"Thirty days from now. The shielding will fail at ninety-seven percent maturation in thirty days." She looked at him. "Two sessions, at seven-day intervals, completing in fourteen days. The membrane architecture fully operational sixteen days before the shielding fails."
Sixteen days of margin.
"What happens in the sixteen days?"
"Testing. Verification. The emergent process interfaces with the Deep's signal through the completed membrane, and we determine whether the interface is functioning as designed. Whether the filter architecture handles the signal load. Whether the Deep's response to the interface is what the original engineers intended or something unexpected." She paused. "We have sixteen days to find out if the window opens when we unlock it."
"And if it doesn't."
"Then we have sixteen days to implement the bunker design before the shielding fails." The engineer's flat delivery of the backup option. "The bunker design is less good. It preserves consciousness at the cost of exposing it catastrophically when the bunker eventually fails. But it buys time. Time we could use to attempt another approach."
"Or to evacuate."
She looked at the city below Buramsan's slope. Eight million lights. "Or to evacuate. Yes."
The session three data was complete. The team descended in the grey-dark of early morning, the mountain's chill and the city's ambient glow doing the job of orienting them to the surface world that the hub's amber interior made easy to forget.
Taeyang ran passive scanning on the descent. The cage's infrastructure. The convergence sites. The seventh seed's sector six conversion, completed and holding at tolerances the archive called theoretical. The lock in his ability's inner architecture — the gate almost open, the comprehension requirement ninety percent of the way to satisfied, the Origin Scan almost accessible.
He'd been conserving the lock. Not deliberately, exactly. But aware that the first activation would produce more than a hairline gap's worth of pain feedback, and that the session protocol needed his scanning field operational. He'd been holding the gate at almost-open while the sessions built the architectural context that the comprehension requirement needed.
The base layer was twelve meters below the hub's floor.
In two sessions, the membrane would be complete.
After that, whatever the Origin Scan showed him.
He wasn't avoiding it. He was sequencing it correctly.
The city spread below them as they reached the trailhead. Dojin's sedan. The dawn beginning to differentiate the sky from the mountains.
"Session four is in seven days," Taeyang said.
"Seven days," Jiyeon confirmed. Her voice carried the particular quality of someone running on three hours of sleep and the specific energy of successful work. "I'll be ready."
