Four days passed. The watcher's response hadn't arrived. The gardener's broadcast hadn't stopped. The operation continued at the pace the timeline demanded: battery maintenance, substrate bond monitoring, integrity assessments at all seventeen sites. The institutional machinery processed its daily inputs — Minsu's reports, the Board's oversight requests, Sera's training rotations, Junseong's scheduling matrix — and the machinery didn't know that the people operating it were carrying a question that made the machinery's purpose uncertain.
Dohyun ran the numbers every morning. Infrastructure integrity at 56.2%. Climbing. The repair trajectory showed 80% in five months if the current rate held. Activation readiness in nine months. The timeline fit inside the twelve-month window with three months of margin.
The numbers were clean. The numbers said the plan was working.
The question was whether the plan should work.
He sat with it the way a soldier sits with a wound that hasn't been treated: aware of it constantly, managing the pain through discipline, functioning around the damage. The question lived in the gap between "can we fire the weapon" and "should we fire the weapon," where military planning met moral calculus and the War Manual was useless.
The team split along predictable lines.
Junseong ran the analysis cold. His notebook had a new section: "Activation Decision Matrix." Two columns. FIRE and DON'T FIRE. Consequences, probabilities, unknowns. The matrix was incomplete because the watcher's data hadn't arrived. Honest about its incompleteness.
Sera defaulted to the position she'd held since the beginning of the operation: protect the people who are alive now. Future threats are future problems. Twelve million people in Seoul who would die in the collection event were real. The dimensional barrier collapse that might happen in sixty or ninety years was theoretical. "We fight the war in front of us," she said at the Wednesday briefing. "We don't sacrifice twelve million people on a probability model."
Seokhwan said nothing at the briefings. He sat in his corner and listened. His silence was a different kind of argument: the argument of a man who'd spent years inside the infrastructure, cutting through the channels, touching the substrate, feeling the geological hum of the architects' work through his blade and his hands and the mana sensitivity that three years of underground labor had grown in him. The infrastructure was real to Seokhwan in a way it wasn't real to anyone else on the team. He'd cut it. He'd felt it break. He knew what it cost to repair and he knew what it would mean to destroy it.
Junho listened and cooked. The meals at Lee's Kitchen continued: breakfast service for the team members who still used the restaurant as an informal meeting point, dinner service for the neighborhood regulars who didn't know the back room contained a geological warfare coordination center. He brought coffee to the briefings. He kept the kitchen clean. He processed the crisis through the work of feeding people, the way a medic processes a battle through the work of treating wounds.
Minhee worked. She built verification models. She cross-referenced the gardener's claim against every piece of infrastructure data the operation had collected. She ran the throughput calculations at every percentage from 90% to 100%, mapping the feedback energy curve, identifying the failure thresholds for each infrastructure component. Her analysis was exhaustive, precise, and ultimately limited by the same gap that limited everything: the watcher's data hadn't arrived.
Soojin worked a different angle. The human one. Ahn Jiseok.
---
"He's alive," Soojin reported at the Thursday briefing. "Ahn Jiseok. Fifty-three years old. B-rank mana ecology researcher. Currently employed at the Haneul Mana Ecology Institute, a private research organization funded by corporate sponsors and government grants. Office in Gangnam."
She put his file on the projector. The institutional record from the Association's personnel database, supplemented by academic publication records and the professional biography from his institute's website.
"Published papers: twenty-seven in the last fifteen years. Topics range from mana ecosystem dynamics to dungeon ecology to long-term mana-density trends. His research is legitimate. Peer-reviewed. Cited by other researchers in the field."
"Anything on the collection mechanism?" Minhee asked.
"Not directly. His published work uses the framework of 'mana ecosystem regulation' without referencing the specific mechanism. But three papers describe phenomena that map to the collection mechanism's effects: periodic mana-density resets observed in geological records, historical correlation between mana-density peaks and ecological disruption events, and a 2024 paper titled 'Self-Regulating Dynamics in Mana-Dense Ecosystems' that describes a theoretical regulatory function that operates on multi-century timescales."
"He published the theory."
"In academic language. Without naming the collection mechanism or the infrastructure. He described the regulatory function as a natural phenomenon. A pressure valve. The papers argue that the mana ecosystem has a carrying capacity, that exceeding the capacity triggers corrective events, and that the corrective events are necessary for long-term ecosystem stability."
"The same argument from Donghwan's journal," Junseong said. "Published in peer-reviewed journals. Available to anyone who reads mana ecology research."
"Nobody reads mana ecology research," Minhee said. "The field has twelve active researchers in Korea. The journals have circulation numbers in the low hundreds. Jiseok's papers are publicly available and functionally invisible."
"By design?" Dohyun asked.
"Unclear. He could be a researcher publishing in an obscure field because the field is obscure. Or he could be someone who wants the information on the record without drawing attention to it." Soojin closed the file. "I also checked his personal records. Property, financial, travel. Standard pattern for a researcher at his level. No anomalies. No connections to the Association beyond his B-rank registration. No connection to Donghwan's estate or the Jongno-gu storage unit."
"He doesn't know Donghwan is dead?"
"Donghwan died in a dungeon break five years ago. The death was reported in the standard Association casualty notice. Jiseok may have seen the notice or may not have. There's no indication that they maintained contact after the encounter described in the journal."
Two men. One a regressor working alone in dungeon sub-levels. The other a scientist working alone in academic journals. They'd met once in a cavern beneath Gwangmyeong and disagreed about whether the world should be saved or pruned. Then they'd gone back to their separate wars: Donghwan to fire the weapon, Jiseok to publish papers that nobody read.
"Do we approach him?" Soojin asked.
The question went to Dohyun. Everything went to Dohyun. The central node that the cell structure required: the person who held the complete picture and made the decisions that the complete picture demanded.
"Not yet," he said. "We wait for the watcher."
---
Saturday. The watcher's response hadn't arrived. The gardener's broadcast continued. The integrity number climbed to 56.8%.
The briefing was small. Dohyun, Minhee, Junseong. The three-person decision-making core that had evolved during Arc 3 when the operation's complexity had exceeded what a single commander could process.
"We're losing time," Junseong said. "The verification process is necessary but it's consuming operational bandwidth. Every briefing for the last week has been dominated by the gardener's claim and the journal's revelation. The repair work continues but the team's focus is split."
"The team's focus should be split," Minhee said. "The gardener's claim, if verified, changes the operational objective. We can't plan for activation without knowing what activation costs."
"We know what not activating costs. Twelve million people."
"We know what activation costs at ninety-three percent. Forty percent channel destruction. We're extrapolating to one hundred percent with a throughput model that hasn't been validated against observed data because one hundred percent has never been achieved."
"The model is the best tool we have."
"The model is a prediction. The watcher's data is observation. The distinction matters."
Junseong looked at Dohyun. "You've been quiet."
He had been quiet. For days. The quiet of a man processing something too large for conversation, the way a computer runs calculations without displaying the screen. The War Manual turning in his head, the twenty-four years of foreknowledge scanning for anything — any event, any fragment, any half-remembered conversation from the first timeline — that addressed the question of infrastructure cost.
Nothing. The first timeline had no ring circuit activation. The infrastructure was never discovered. The question was new.
"I've been thinking about Donghwan," he said.
"The previous regressor."
"He faced this choice. Jiseok told him the weapon would destroy the channels. He fired anyway. At ninety-three percent, not one hundred, but he knew the damage would happen and he fired."
"Because he was alone," Junseong said. "No team. No institution. No verification capability. He had his foreknowledge and his judgment and a weapon that might work. He fired because the alternative was doing nothing while the collection event killed millions."
"He made the right call," Sera's voice from the doorway. She'd arrived without announcing herself, the combat habit of entering a space quietly to assess it before committing. She stood in the frame, arms crossed. "He saved the world. The channels burned. The world survived. That's the math."
"The math at ninety-three percent," Minhee said.
"The math at any percentage. You fire the weapon, people live. You don't fire the weapon, people die. Everything else is a cost that you pay after survival."
"Unless the cost of survival is extinction one generation later."
"One generation later is one generation later. The people who are alive now have the right to be alive. You don't sacrifice the living for the theoretical future."
Minhee looked at her. The look of two women who respected each other and who were arriving at the same question from opposite directions: one from the data, one from the people the data represented.
"The theoretical future contains people too," Minhee said. "If the dimensional barriers collapse, the people who die aren't theoretical. They're the children and grandchildren of the people we're saving."
"Then the children and grandchildren rebuild the weapon."
"In ninety years. With a fractured infrastructure and no architects' knowledge."
"With whatever they have. The same way we're fighting with whatever we have." Sera stepped into the room. Sat down. The combat stance relaxing into the operational posture of someone who'd said her piece and was now part of the conversation instead of adjacent to it. "You're asking me to weigh twelve million certain deaths against a probability model of future deaths. I'm not smart enough to do that math. I'm not sure anyone is."
The room was quiet. The three-person core plus one. The four of them around a table that had held larger groups and smaller groups and that was now holding the weight of a question that no group was large enough or small enough to answer properly.
"We verify," Dohyun said. "The watcher's data arrives and we verify. If the gardener's claim is true — if one hundred percent activation destroys the infrastructure permanently with no rebuild path — then we have a different conversation. If the claim is partially true — if the damage is severe but recoverable — then we have a cost we can plan for."
"And if the watcher's data is ambiguous?"
"Then we decide with incomplete information. The way every decision in this operation has been made. The way every decision in every war has been made."
Junseong wrote something in his notebook. One line. He showed it to Dohyun.
*Pocheon. Full descent. The relay isn't enough.*
The relay protocol was limited. Binary confirmations. Intensity-coded responses. The bandwidth of a signal bounced through a repeater two hundred kilometers into the earth. The question they needed answered was too complex for binary.
"The relay isn't enough," Dohyun agreed. "We need the watcher's full dataset on infrastructure response to ring circuit activation. The Cycle 4 data. The projections. The specifications the architects wrote for full-power operation."
"That requires a Pocheon descent," Minhee said.
"A Pocheon descent with the Association's security escort. The institutional resources that we didn't have before."
"The dungeon ecology has been fully compromised by the gardener. The boss creature is turned. The garrison is gone."
"Yeonhwa knows the path. The watcher's chamber is still accessible. The crystal is below the gardener's colonization depth."
Minhee looked at her laptop. The relay data showing the watcher's steady pulse, the heartbeat of an entity waiting in the dark for the allies it had agreed to help. The pulse was weaker than before. The gardener's dungeon ecology pressing closer to the crystal's threshold. But the pulse was there.
"I'll organize the descent," she said. "Full security escort. Association resources. Yeonhwa leads."
"Fast," Dohyun said.
"This week."
The briefing ended. Junseong left with his notebook. Sera left with the expression of a woman who'd spoken her position and who would defend it when the data arrived. Minhee stayed, organizing the Pocheon operation on her laptop, the methodical planning of a woman who preferred to verify before she decided and who was not willing to let the verification wait.
Dohyun stood at the operational board. The numbers. The timeline. The gardener's broadcast pulsing beneath the building, beneath the city, beneath everything.
THE WEAPON DESTROYS MORE THAN THE COLLECTOR.
Trust nothing, he thought. Verify everything. Decide with the best information available and live with the consequences of what the information didn't show.
The creed of a soldier in a war where the enemy might be right.
