Rowan's model was seven pages.
Not a written report — something more structured than that. A chart, a timeline, a set of probability curves, the specific kind of document that Rowan produced when he'd been thinking about something for weeks and was now ready to show someone else what he'd found. Kael sat at Rowan's worktable that evening and read through it twice while Rowan sat at the far side of the table running secondary calculations and not commenting.
The first page was the baseline: what Kael remembered of the original timeline through the same period. Dungeon first-clears, Guild formations, Hunter Association decisions, key individual events. Rowan had compiled this from their conversations over the past two months, cross-referencing against current public records to identify divergence.
The second page was the divergence record: what had actually happened versus what Kael had anticipated.
The list was longer than he'd expected.
"Nine major divergences," he said.
"Eleven, if you include the modification program exposure and the talent designation access." Rowan's voice came from behind the laptop. "I classified those at lower significance because they represent information differentials rather than event changes. You know things that aren't supposed to be knowable yet. That has effects that aren't fully traceable."
He looked at the list. The Voss operation consumed three entries — the scan, the disruption, the intervention. The dungeon he'd entered that had the anchor marking consumed two — the different map, the territorial mob that hadn't existed in the original timeline. Two more for what Rowan called "interpersonal divergence" — Elara's sister's fate, and whatever knock-on effects Lena's natural awakening instead of Voss's modification would produce. Then smaller things: the Association chamber assessment, the early talent designation.
"What's the Watcher threshold," he said. "From the model."
Rowan was quiet for a moment. "I don't have reliable data for that. What I have is the second regressor's documented divergence rate compared to yours, extrapolated from Voss's anomaly signature research." He turned the laptop. "The second individual — the one who triggered the Watcher at month eleven — was operating at approximately double your divergence rate. If that ratio holds, you have a theoretical buffer." He looked at the model. "But Voss's note about the fourth individual complicates the model. The system-level correction may operate on a different variable entirely. Peak divergence in a window rather than cumulative total."
"Peak-versus-cumulative."
"If it's cumulative, you're fine. Current pace, you're generating sustainable divergence with adequate spacing." He tapped the laptop. "If it's peak-based — if one six-week window of high-density change is sufficient to trigger the correction — then the Voss operation window may have already approached the threshold."
Kael looked at the probability curves on page four. The curves spread wider toward the future, the error bars widening as the model extended further. Rowan's model was honest about what it didn't know, which was most things.
"What's your recommendation," he said.
Rowan closed the laptop. Folded his hands. The particular stillness he got when he was about to say something directly and was bracing for the landing.
"Two things. First: reduce the pace of significant interventions for the next sixty days. Allow the timeline to stabilize. Let the divergences from the last eight weeks become background rather than active interference." He paused. "This means pausing anything you've been planning that would constitute a major change."
"And second."
"Second." A pause. "What are you planning for Marcus Thorne?"
Kael didn't answer immediately. The question sat in the room.
Rowan continued: "I know you have a plan. I know it involves his awakening class. I don't know the specifics, but I've noticed you've visited the Church of the Awakened district twice this week for no stated reason, and you've asked me three separate questions about class-disruption mechanics that weren't relevant to the Lena operation." He met Kael's eyes. "If the Marcus plan involves another intervention before the timeline stabilizes—"
"I know."
"Do you?"
Kael looked at the divergence model.
He looked at the probability curves spreading wide toward an uncertain future.
"Sixty days," he said.
"That's the recommendation." Rowan's voice was careful, not urgent. Precise. "It gives the current divergences time to stabilize before adding new ones. It gives you time to rebuild the training pace, which is currently behind." He looked at the model. "The Marcus problem isn't going away. His awakening window is the next two to three months. There's time to plan rather than rush."
"His awakening window closes."
"It closes for the optimal disruption point. But disruption isn't the only approach." He looked at Kael. "I've been thinking about the Marcus problem independently." This was a thing he did — worked problems in the background and brought the results without being asked. "In the original timeline, Marcus became dangerous because he had a specific class, a specific resentment dynamic, and a specific set of social relationships that Dorian could exploit. You've been focusing on the class variable. But the resentment and the social relationships are also variables."
Kael looked at him.
"The resentment in Marcus is real," Rowan said. "I've run a surface read on his social pattern from Association public records and Church registration documents. He's talented, recognized within the Church but not outside it, consistently overlooked when broader platforms emerge. The dynamic exists independent of what class he gets." He folded his hands. "If the class disruption works and he doesn't get the healer path — what does he become? What class uses his mana perception ability without the healing orientation? We don't know, because we've never modeled it." He looked at Kael. "What if the disruption produces something more dangerous rather than less?"
The question sat there.
Kael had been assuming the disruption would neutralize Marcus. Remove the divine healer class, remove the tool Dorian would eventually use. But Rowan's point was correct — the class was one variable among three, and the others remained.
"He becomes something else," Kael said slowly. "Not a healer. Something with his natural mana perception ability, without the healing framework."
"Yes."
"Could be worse."
"Could be significantly worse." Rowan looked at the probability curves. "The healer class comes with institutional structure — Church oversight, ethics frameworks, formal training protocols. It constrains the expression of his ability into a specific shape." He paused. "Without that structure, whatever class his natural ability produces is unconstrained. And someone with Dorian Vex's specific talent for finding leverage and applying it to unconstrained ability—"
"Would have more options, not fewer."
Rowan didn't answer that, which was his way of confirming it.
Kael sat with the model and the new shape of the problem.
He'd been thinking about this wrong. He'd been treating Marcus as a variable to minimize, a class to block, a resource to deny Dorian. That was the right framework if Marcus was only a piece. But Marcus was a person with interlocking components — ability plus resentment plus social dynamics — and removing one didn't remove the others. It just changed how they expressed.
"You're saying I should consider whether leaving the healer class intact is strategically superior."
Rowan was quiet for a moment. "I'm saying the problem is more complex than 'block the healing class.' I don't have a definitive recommendation. I'm flagging the assumption."
Kael looked at the model for a long time.
"Table it," he said finally. "Sixty days. No Marcus intervention until we've built the analysis properly." He looked at the training timeline Rowan had included on page six — the daily conditioning plan, the projected E-rank window, the dungeons he needed to clear. "I need to get back on schedule."
Rowan nodded. Opened the laptop again. "I'll update the schedule for tomorrow. You have a training window at 0600, and the Thornback corridor's northern branch hasn't been first-cleared yet — the main floor was claimed but the northern branch maps as a separate clear."
He'd lost the main floor. But the branch was still there.
"0600," he said. "I'll be there."
He stood. Moved toward the door.
"Kael."
He looked back.
Rowan was looking at the divergence model, not at him. "The Voss information. The four previous regressors." He turned a page. "One of the failure modes is insufficient information at a critical juncture." He looked up. "You walked away from her offer at the canal."
"I chose Lena."
"I know." He paused. "I'm not re-arguing the choice. I'm noting that the gap in your information about how regressive displacement can fail is now the most significant strategic vulnerability you have." He looked back at the model. "Whatever you decide about working with Voss — the information she has about the failure modes is not optional. You need it."
Kael stood in the doorway.
He'd been telling himself he'd decide about Voss later. Assess when he had time. But Rowan's analysis had the quality of something being said because the moment for delaying was ending.
"I know," he said.
He went to bed and ran the divergence model figures through the strategic overlay of his forward memory — everything he remembered about the next ten months, mapped against eleven documented divergences and four failure templates from people who'd stood in the same position and burned out, been erased, or simply vanished.
The training schedule was at the center.
The Marcus problem was paused but not gone.
And Iris Voss was waiting for him to decide what she was worth.
He closed his eyes and let the planning run, the way it always ran now, the background process that never fully switched off. Somewhere in Ravenscrest, Lena Winters was sleeping and the light in her hands was learning what it wanted to be. Somewhere, Marcus Thorne was two months away from an awakening Kael hadn't decided how to handle.
And somewhere, in a research space on the fourth floor of a converted commercial building, Iris Voss was filing a new document in a folder labeled with a name he'd never seen.
He fell asleep thinking about the fourth regression.
The clean disappearance. Like it had never happened at all.
