Dohyun told them at 08:00.
Not just the Intelligence cell. Junseong. Junho. Seokhwan. Everyone who held operational responsibility. He'd spent the hours between 03:00 and dawn running through the War Manual's contents the way a mechanic runs through an engine looking for parts that might be salvaged, and by 07:30 he'd decided that compartmentalizing the regressor evidence would do more damage than sharing it. The finding changed the foundation. The foundation supported everyone.
He read the translated entry aloud. The room listened.
When he finished, the paper sat on the table between them. Entry 7. Year 2, Month 4. The words of someone who'd been here before.
Junseong spoke first. He'd been listening with his notebook open but hadn't written anything. The pen was on the table beside the book.
"You've been making decisions based on foreknowledge for three years," he said. "You predicted the Gangnam break. You identified future S-rank hunters before they developed. You found the infrastructure before anyone else. Your entire operational approach is built on the assumption that you know what's coming." He picked up the pen. Held it between two fingers without writing. "Now you're telling us that the foreknowledge is a copy. That someone else carried a similar version and used it to run an operation that failed. Possibly multiple times."
"That's what I'm telling you."
"So how do we know any of your predictions are still accurate?" The question came without heat. Clinical. "Your Gangnam numbers were two hundred. The previous regressor's were eighty-three. Different versions of the same event. Which version describes what will happen in this cycle?"
"I don't know."
Junseong put the pen down. "Good. That's the right answer."
Junho was leaning against the counter. Arms crossed. His bruises from Pocheon had darkened to the purple-black stage, visible above the collar of his shirt. "So the man who knows everything is admitting he doesn't know everything. What changes?"
"The methodology changes," Minhee said. She'd been quiet. Organizing. The way she organized before she spoke, building the argument in her head before laying it on the table. "Dohyun's foreknowledge has been our primary intelligence source since the operation began. We've treated it as a prediction engine. When Dohyun says an event will happen, we plan for that event. When he identifies a threat, we prioritize that threat."
"And it's worked," Junho said. "Mostly."
"Mostly. The Gangnam prediction was accurate in kind but not in detail. The gardener's behavior was predicted but the speed was wrong. The Bucheon pressure was anticipated but the acceleration wasn't. Each prediction has been correct in its general direction and incorrect in its specific parameters."
"Which is what you'd expect from intelligence gathered in a different timeline," Junseong said. "The broad patterns repeat. The details change. It's like running the same war from a different starting position."
"Exactly. So the approach isn't to discard the foreknowledge. It's to stop treating it as a script." Minhee pulled her laptop from her satchel. Opened a spreadsheet she'd clearly been building since last night. "I've catalogued every prediction Dohyun has made since the operation began. One hundred and twelve specific claims about future events, people, or conditions. Of those, seventy-four have been verified as broadly accurate. Twenty-three were incorrect. Fifteen are unresolved."
"Sixty-six percent accuracy," Taeyang said. He'd been running the numbers in his head.
"Sixty-six percent. Which is better than any intelligence source in military history except direct observation. But it means one in three predictions is wrong. And we can't tell in advance which third."
"Cross-referencing," Junseong said. "Treat foreknowledge as one source. Verify against current intelligence before acting. Sensor data. The watcher's records. Field observation. Real-time assessment. Don't throw the manual away. Just don't read from it with your eyes closed."
Dohyun sat at the table and listened to his team dismantle the foundation of his advantage and rebuild it as something sturdier. Three years of operating as the man who knew the future, the Prophet, the person whose predictions shaped every decision. And now his team was telling him, correctly, that the predictions were a tool, not a religion.
"The verification protocol adds time to decision-making," Dohyun said. "Some decisions are time-critical. Bucheon pressure spikes. Gardener agent incursions. If we verify every prediction before acting—"
"Not every prediction," Minhee said. "We build a confidence tier. Tier one: predictions that have been verified in this cycle by independent data. These get acted on immediately. Tier two: predictions that match the general pattern but haven't been independently confirmed. These get flagged for verification before major resource commitments. Tier three: predictions that conflict with current data or that relate to events the previous regressor's entry suggests went differently across cycles. These get treated as hypotheses, not facts."
"Who classifies which tier?"
"I do. Intelligence cell function. I cross-reference each prediction against our current data set and assign a confidence level. The tier classification is updated as new data arrives."
The spreadsheet on her screen showed the framework. One hundred and twelve predictions, each with a date, a description, a verification status, and a blank column for tier assignment. She'd been working on this since the decoded entry. The kind of person who responded to a destabilized foundation by redesigning the building.
"Approved," Dohyun said.
---
Taeyang delivered the operational update while Minhee began tier classifications at the counter.
"Battery 3 at Anyang is back online. Randomized frequency active. Haejin's team confirmed stable resonance lock at six point two times natural regeneration rate. Lower than the matched-frequency approach, but the counter-frequency risk is eliminated for that site."
"Other batteries?"
"All remaining batteries are being converted to randomized frequency protocols this week. The Repair cell is handling the recalibration schedule. Estimated completion: five days for all sites."
"Bucheon?"
"Pressure at sixty-three percent. Declining at one point one percent per week. The western artery repair is at eighty-two percent of target, the highest of any site. The battery efficiency cascade that Yeonhwa identified is continuing. The longer the batteries run, the faster the repair proceeds."
Sixty-three percent. Down from the seventy-seven that had nearly broken the team a month ago. The containment war was being won. The pressure was declining. The batteries were working, even at reduced efficiency from the randomized frequencies.
"Infrastructure integrity?"
"Fifty-four percent. Up from forty-six when we started the Bucheon operations. The rate of improvement is accelerating as more sites come online."
Fifty-four. Twelve points of repair in a month. At this rate, the primary channels would reach functional levels in five to six months. Add the secondary conduit backup, and the ring circuit could theoretically fire within the year.
The numbers were good. Better than any previous chapter of the operation. Better, possibly, than anything the previous regressor had achieved.
"In Cycle 2, the previous regressor reached sixty-seven percent integrity before the counter-frequency campaign destroyed the batteries," Dohyun said. "We're at fifty-four with counter-frequency countermeasures in place. If their sixty-seven was the highest any previous cycle achieved, we're on pace to exceed it within two months."
"Because of the randomized frequencies," Taeyang said.
"Because of the randomized frequencies. Because of the cell structure distributing the operational load. Because of Minhee's throughput model confirming the backup pathway. Because of the watcher." Dohyun looked at the room. At the people. "Because the previous regressor, whoever they were, worked alone or nearly alone. They carried the foreknowledge and ran the operation and made the decisions and took the casualties by themselves. And the gardener beat them. Cycle after cycle."
"We don't know they worked alone," Minhee said from the counter. Precision. Always precision.
"The entry refers to 'my early-warning intervention.' 'I will prioritize.' Singular first person throughout. No mention of a team, a cell structure, or organizational assets. The tone is a field commander reporting to themselves."
"A field commander reporting to themselves," Junseong repeated. He picked up his pen. Finally wrote something. "Lone operator model. Single point of intelligence, single point of decision, single point of failure. The same structural vulnerability I identified in your operation three weeks ago, except the previous regressor never fixed it."
"And I did. Because you told me to."
"Because you listened." Junseong closed the notebook. "The previous regressor's failure mode wasn't lack of foreknowledge. It was lack of organization. They knew what the gardener would do. They knew how the infrastructure worked. They knew about the secondary conduits. But they tried to operate the entire repair campaign as an individual, and the gardener has eight hundred years of experience beating individuals."
Seokhwan stirred in his corner. He'd been cleaning his blade. The motion stopped.
"The gardener modifies people one at a time," Seokhwan said. His voice was the same quiet register he always used, the tone of someone who'd been the gardener's tool and who understood the enemy from the inside. "One agent. One modification. One target. The gardener operates on individuals because that's how its modification architecture works. It can't modify a team. It can't modify a cell structure. It can install cutting technique in one person's mana profile, but it can't install it in an organization."
"The organization is the countermeasure," Junseong said.
"The organization is something the gardener hasn't faced before." Dohyun stood. Walked to the operational board. The new board, with the cell sections and the pressure readings and the repair percentages. "In every previous cycle, the regressor found the infrastructure and tried to fix it alone. The gardener learned to beat that approach. Counter-frequency against the batteries. Modified agents against the keystone. Intelligence-gathering agents against the repair protocols. Every capability the gardener has developed is a response to an individual operator."
He turned to the room.
"We're not an individual operator. We're four cells with distributed leadership, compartmentalized intelligence, redundant decision chains, and a succession plan. The gardener's counter-frequency attack works against a single battery protocol because one person chose one frequency. Our randomized protocols use different frequencies at every site because an engineering team designed a system, not a single operator picking numbers."
"The gardener's modification architecture targets individuals," Taeyang said, following the logic. "One person, one modification, one behavioral pattern. But an organization with rotation protocols, buddy-system monitoring, and regular mana scans doesn't rely on any single individual's integrity."
"The intelligence-gathering modification worked because a single survey specialist uploading data to a centralized server created a single pipeline the gardener could exploit. Under the new Security protocols, data doesn't touch any centralized server. The pipeline doesn't exist because the architecture doesn't allow it."
The room was processing. Not the regressor revelation anymore. The implications of the revelation applied to their operational advantages. The shift from "we might fail because someone else did" to "we might succeed because we're doing something different."
"This doesn't mean we're safe," Junseong said. The counterweight. The voice that never let optimism run unchecked. "The gardener learns. It adapted to the previous regressors' individual approaches by developing counter-individual capabilities. If it faces an organizational approach for the first time, it will adapt. It will develop counter-organizational capabilities. We don't know what those look like because they haven't existed before."
"How long does adaptation take?"
"In this cycle, the gardener went from basic cutting modifications to four-layer intelligence-gathering architecture in approximately six weeks. If organizational countermeasures are a new problem, the adaptation timeline might be similar."
"Six weeks," Dohyun said. "Then we have six weeks of organizational advantage before the gardener develops a response. Six weeks where our structure is something it hasn't seen and can't counter."
"Approximately. Assuming the adaptation rate is constant, which it probably isn't."
Junho pushed off the counter. "So we have a window. Six weeks where the team structure is our edge. What do we do with it?"
The question hung in the room. Dohyun looked at the operational board. The numbers. The repair percentages climbing. The pressure dropping. The battery network expanding. The infrastructure recovering.
"We finish the repair," he said. "We push the integrity past sixty-seven percent, past the previous cycle's peak, past the point where the gardener has historically been able to stop the operation. We get the primary channels to functional levels while the organizational advantage holds. And we prepare for whatever the gardener develops next."
"And the watcher?" Minhee said. She'd stopped her tier classification. This was the question she'd been waiting to ask. "The remaining log entries are in the watcher's crystal. If we can read them, we might find out what the previous regressor's secondary conduit approach produced. Whether it succeeded or failed. What the gardener's counter-organizational adaptation looks like in a future cycle."
"The watcher is our next objective after the repair timeline stabilizes. Pocheon return operation. The gardener contested our last visit, but the watcher held control. If we go back with a plan for extended contact, a security protocol for the gardener's interference, and the organizational support to execute both simultaneously—"
"Then we get the answers the previous regressor left for us," Minhee said.
"For whoever came next." Junseong's voice. He was looking at the decoded entry on the table. The paper that someone had written thirty years ago and stored in geological crystal. "They wrote those entries knowing they wouldn't be the ones to read them. They wrote them for a stranger in a future cycle who might do what they couldn't."
He picked up the paper. Read the last line.
"'I am running out of cycles.'" He set it down. "They ran out. We haven't. That's the other difference between us and them."
The briefing ended. The team dispersed to their cell assignments. Junseong took the Containment cell to Bucheon for the scheduled clear. Minhee returned to her tier classification. Taeyang went to the monitoring station. Seokhwan headed to the Suwon rotation.
Dohyun stood at the operational board. Alone. Lee's Kitchen in the mid-morning quiet, the kind of silence that a restaurant has when it's between meals, the space holding its breath between one use and the next.
He took the decoded entry from the table. Read it one more time. The handwriting of another soldier. Another commander. Another person who'd died and come back and tried and failed and left a message in the dark for someone they'd never meet.
He pinned the paper to the operational board. Next to the cell assignments. Next to the pressure readings. Next to the note that said *Watch the architect.* Next to the throughput calculation that said seven percent margin.
The board held the entire operation. The plans and the people and the numbers and the warnings. And now it held the words of someone who'd stood in the same position and who'd wanted the person who came next to know one thing:
This has been tried before. Learn from it. Do better.
Dohyun picked up a marker. Wrote beneath the pinned entry, in the field notation that only he could read:
*Cycle 5. Organizational approach. Team of twelve. Four cells. Not alone.*
*Not this time.*
