Minhee took three days with Jiseok's data.
Three days of the eleven-month timeline. Three days of the repair operation continuing at Phase 2 pace, the batteries humming, the integrity climbing, the institutional machinery processing its inputs while the Intelligence cell's lead analyst sat in her office with Jiseok's geological surveys and crystal-lattice analysis and notation translations and ran them against every piece of verified data the operation possessed.
Dohyun didn't rush her. Rushing Minhee produced faster answers and lower confidence levels, and the question they were asking required the highest confidence level the operation could achieve.
She presented at the Friday briefing. Full team. Lee's Kitchen. The back room with the board and the chairs and the familiar geography of decisions made in a restaurant that smelled like coffee and sesame oil.
"Jiseok's calibration analysis is partially correct," she said.
The room waited.
"The keystones do have variable power settings. The architects' notation, as encoded in the crystal lattice of the substrate around each keystone site, includes calibration parameters that specify different activation intensities. Jiseok identified four settings: twenty-five percent, fifty percent, seventy-five percent, and one hundred percent. His translation of the notation is different from mine but the underlying data matches."
"The weapon has a dial," Sera said.
"The weapon has preset configurations. Not a continuous dial. Four fixed settings. Each setting determines the amount of feedback energy the ring circuit generates. At one hundred percent, the energy overwhelms the collection mechanism by a factor of four point one and destroys the infrastructure. At fifty percent, the energy overwhelms the mechanism by a factor of approximately one point two."
"One point two. That's a twenty percent margin."
"A twenty percent margin on a system that has never been tested at any setting. The simulation data from the architects covered one hundred percent. The lower settings are specifications without test data."
"Jiseok's claim about modulated activation," Junseong said. "Fire at fifty percent. Reduce the collection but don't prevent it. Hundreds of thousands die instead of twelve million. The infrastructure survives."
"That's his claim. My analysis supports the energy calculations. At fifty percent, the feedback energy is within the primary channels' structural tolerance. The channels survive. The secondary conduits survive. The substrate bonds hold. The infrastructure takes damage but doesn't fail."
"And the collection mechanism?"
"At fifty percent, the mechanism is weakened. Not destroyed. The collection proceeds but at reduced intensity. Jiseok's casualty model projects two hundred thousand to five hundred thousand deaths in the Seoul extraction zone, depending on how the mechanism responds to partial suppression."
Two hundred thousand to five hundred thousand. The numbers hung in the room like smoke. Not the clean zero of full activation. Not the catastrophic twelve million of no activation. A middle number. A compromise measured in human lives.
"That's Jiseok's model," Sera said. "His projection. His numbers. What do our numbers say?"
"Our numbers are limited," Minhee said. "The watcher's data covers the previous collection event eight hundred years ago — at full intensity, with no weapon activation. The collection killed the architects' civilization. We don't have data on a partially suppressed collection because it's never been attempted."
"So the two hundred thousand to five hundred thousand is a guess."
"It's a model-based projection with no observational validation. The range is wide because the uncertainty is large. At the lower bound, the modulated activation suppresses the collection to a manageable level. At the upper bound, the collection adapts to the suppression and proceeds at near-full intensity despite the partial activation."
"The collection adapts?"
"The collection mechanism is described in the architects' notation as an 'external harvesting entity.' Entity, not process. Entities adapt. Processes don't." Minhee paused. Let the distinction register. "If the collection mechanism is an entity — something with behavior that responds to stimuli — then a partial activation might not produce a predictable reduction. The entity could route around the suppression. Could target areas outside the ring circuit's coverage. Could intensify in response to the weapon's interference."
"Could," Junseong said. "Not will."
"Could. We have no data on the mechanism's adaptive behavior because nobody has ever tried to partially suppress it."
The room digested that. The third option: fire at fifty percent, hope the mechanism reduces predictably, accept hundreds of thousands of deaths, preserve the infrastructure. The option that was elegant in theory and terrifying in practice because the practice involved an entity that might not cooperate with the theory.
---
Dohyun let the team talk for an hour. The positions formed. The lines drew.
Sera: "Full power. Zero compromise. The mechanism dies. The people live. All of them."
Junseong: "The modulated option preserves the infrastructure and the weapon for future threats. The full-power option eliminates the current threat and removes all defense for ninety years."
Sera: "Ninety years of no weapon is better than five hundred thousand dead today."
Junseong: "Five hundred thousand is the upper bound. The lower bound is two hundred thousand. And the weapon survives."
Sera: "The lower bound is a projection from a scientist who's never tested his model. The actual number could be higher. Could be five million. Could be twelve million if the mechanism adapts."
"Could be fifty thousand if the suppression works better than projected," Junseong said. "The uncertainty cuts both ways."
"The uncertainty is exactly the problem. Full power has no uncertainty. The mechanism is destroyed. The people survive. The infrastructure is lost. Those are known costs."
"Known costs with unknown long-term consequences."
"Long-term consequences that the next generation manages with the rebuilt weapon."
"If the rebuild happens in time."
The same argument from the storage unit journal. The same positions that Donghwan and Jiseok had taken. The same irreconcilable tension between the certain present and the uncertain future.
Minhee stayed neutral. She presented data. She didn't advocate. The Intelligence cell's discipline: provide the analysis, let the decision-makers decide.
Seokhwan was quiet. Thinking. Processing the way stone processes pressure: slowly, deeply, with the weight of three years spent inside the infrastructure that the decision would save or destroy.
Junho: "How do you choose between killing nobody today and maybe killing everybody tomorrow?"
Nobody answered that. Because the answer was: you can't. The choice was between different kinds of not-knowing. Different kinds of gambling with lives that weren't yours to gamble with.
"Dohyun," Junseong said. "Commander's call."
---
He'd made the call once already. Full power. Zero compromise. The decision from two weeks ago, made by ten people in this room, unanimous, clean.
He'd made it before Jiseok. Before the variable power settings. Before the third option that preserved the infrastructure and cost hundreds of thousands of lives.
The decision he'd made had been based on incomplete information. The decision wasn't wrong. It was made with what they had. Now they had more.
"I need to say something," he said. "About the cost."
The room waited.
"Two weeks ago, I told you we fire at full power. Save the people who are alive. Deal with the next sixty years after we survive this one." He was standing at the back. The same position. The same board behind the team's heads. "I said that because I believed the choice was binary. Fire or don't fire. Save everyone or lose everyone."
"It was binary," Sera said. "Until Jiseok."
"It was binary because the War Manual didn't know about variable settings. Nobody in any cycle tested the lower settings because nobody had the time or the knowledge."
"Jiseok investigated them because he had twenty years and a different question. Not 'how do we fire the weapon.' 'How do we manage the ecosystem.'"
"What are you saying?" Sera's voice was tight.
"I'm saying I was wrong about the cost."
The room went still.
"I told you the choice was: fire at full power, lose the infrastructure, save everyone. I presented it as a sacrifice. The architects' intent. A clean trade. And it is a clean trade if the only options are one hundred percent or zero percent."
"They're not the only options."
"They're not the only options. Fifty percent is an option. Seventy-five percent is an option. Each option has a different cost in lives and a different cost in infrastructure. And I didn't present those options because I didn't know they existed."
"You didn't know because the information wasn't available."
"I didn't know because the War Manual doesn't cover this scenario and I stopped looking for alternatives when the Manual's answer seemed sufficient." He took a breath. The admission that cost him. The acknowledgment that the twenty-four years of foreknowledge had become a cage as well as a guide, the certainty of the Manual's answers preventing him from asking questions the Manual hadn't considered. "The Manual said fire the weapon. I said fire the weapon. The team said fire the weapon. The decision was unanimous because the decision was based on the same incomplete picture."
"The picture is still incomplete," Minhee said. "Jiseok's model is unvalidated. The variable settings are untested. The collection mechanism's adaptive behavior is unknown."
"The picture is more complete than it was. And the additional information changes the decision space."
"Does it change the decision?" Junseong.
The question. The one that mattered.
Dohyun looked at the board. The numbers. The integrity percentage climbing. The timeline running down. Eleven months. The repair operation approaching the critical path's activation threshold. The weapon being prepared to fire.
"I don't know," he said. "That's what I'm wrong about. Not the decision to fire. The certainty that the decision was obvious."
Sera pushed off the wall. Three steps into the room. "The certainty was the only thing that made the decision bearable. You're telling us that the clean trade — fire the weapon, save everyone, lose the infrastructure — might not be the right trade. That there's a trade where fewer people die in the collection and the infrastructure survives. That the right answer might be somewhere between zero deaths and twelve million."
"Yes."
"And that somewhere between zero and twelve million is a number that means hundreds of thousands of real people in real apartments in real neighborhoods in Seoul die because we chose to fire at fifty percent instead of one hundred."
"Yes."
"And you want us to make that choice."
"I want us to have the information to make whatever choice we make. The previous decision was made without this information. We owe the decision a better picture than what we had."
The room was quiet for ten seconds. Twenty. Thirty.
"Verify the variable settings," Sera said. Her voice had changed. Not the combat register. The register of a woman watching clean certainty erode under information she couldn't dismiss. "Verify everything Jiseok claims. And then we decide."
"Together," Dohyun said.
"Together. Because you were right about one thing at least."
"What?"
"That nobody should make this choice alone."
She walked out. The door closed behind her with the precise force of someone who was not slamming it and who wanted him to know she was not slamming it.
---
The rest of the team left in ones and twos. Minhee stayed.
"The verification will take time. Another Pocheon query. Days. Maybe a week."
"The time isn't the problem," Dohyun said.
"No. The problem is what the verification produces. Full power: zero deaths, no infrastructure. Modulated power: some deaths, preserved infrastructure. The choice between zero and some is a choice between types of responsibility."
"I know."
"Which weight is heavier?"
He didn't answer. The names. The names were always heavier. The infrastructure was engineering. The names were people.
"Send the verification query tonight," he said. "Full analysis of the variable power settings. The collection mechanism's probable response to partial activation."
She packed her laptop. Left. The restaurant empty. The board on the wall.
Dohyun sat at the table. He'd been wrong. Not about the decision to fire. About the completeness of the picture. About the assumption that the War Manual's framework — binary choices with binary outcomes — was the right framework for a problem with more settings than on and off.
His phone buzzed. His mother.
*Saturday. 8:00. The potatoes are ready.*
He typed: *I'll be there.*
The potatoes. Saturday breakfast. His mother's apartment eight hundred meters from the Mapo gate, inside the collection mechanism's extraction zone, inside the number that the decision would determine.
His mother was in the number. Every person he knew was in the number. The number wasn't abstract. The number lived in apartments and ate breakfast and texted about potatoes.
He put the phone down. Looked at the board. The numbers that were clean and the numbers that weren't.
Eleven months. The weapon being prepared. The dial he didn't know existed now visible on the control panel of a machine that saved the world or managed its pruning or something in between that nobody had ever tried and that nobody could predict.
The War Manual had no guidance.
He was on his own.
