Castellan ran the controlled combat integration for fifty minutes. By the end, the wound site had remained quiet and the channel architecture had stayed in the productive register the entire time, which she noted with the expression she used for things that surprised her in minor ways.
"The wound response is better than I expected at this intensity," she said, making the note. "The healing integration is holding under full channel load."
"Good."
"It's good. It's also not an excuse to push harder next session." She closed the tablet. "The left cluster still reads slightly compressed. One more week before we move to the unrestricted intensity range."
He accepted this. One more week.
He was out of the training building by 1545 and at his desk by 1610. The Rowan call was scheduled for 1700. He used the fifty minutes to run the message queue and think about the access code.
The research library in the Association's annex had two tiers: general access, which Fenra's research program ID covered, and archive access, which required either a research authorization or a practitioner registration that had been cross-linked to the annex's evaluation systems. Marcus's assessment practitioner code was registered through the annex evaluation center. Whether that cross-link covered archive access was what Rowan was checking.
If it did, the framing was straightforward: Marcus needed to review the project approval documentation as a research support exercise for his practitioner development. The meeting minutes Kael wanted were in the archive because they predated the current evaluation filing system — an eight-month-old administrative gap that nobody had digitized yet.
If it didn't, he needed another angle.
Rowan called at 1658.
"The access code covers archive access on a restricted basis," Rowan said. He sounded like he'd been running searches since the morning call. "The restriction is that archive access through the practitioner registration requires a documented research justification attached to the access request. The justification gets logged."
"Which means Marcus's name is attached to the access if anyone looks."
"Yes. The log isn't public — it's internal. But if Crane's network has contacts in the annex administration—"
"They might see the access." He thought about this. "How long is the log retention."
"Standard is ninety days." A pause. "If the external oversight committee establishes investigation status within the ninety-day window, the log becomes protected under the investigation's evidence chain. Anyone accessing it after that point would be interfering with an active investigation."
"So if Marcus runs the archive access in the next ten days and the committee's investigation protection kicks in before ninety days—"
"The log is protected before it becomes useful to anyone looking." A pause. "That's a tight sequence but it holds."
"It holds." He looked at the window. "What's the research justification framing for the archive access."
"I've drafted one." He heard Rowan pulling something up. "Marcus is doing an evaluation framework study on historical talent assessment methodologies used by the Association's research division. The project approval meeting minutes fall under that scope — they include notes from evaluation committee members about the methodological basis for the Fenner program approval. Legitimate research territory for an assessment practitioner."
"Is it the kind of research Marcus would actually do."
A pause that was Rowan genuinely considering this rather than deflecting. "It's adjacent to his consent framework work. The consent issues he's been addressing are partly about what assessment practitioners are supposed to document — and the Fenner program approval has documentation standards that differ from current practice. He could credibly argue he found the discrepancy and wanted to understand the historical reasoning." Another pause. "It's a real question."
"It's a real question." He looked at the ceiling. "That's what we give him."
"You'll need to talk to him."
"Monday session." He paused. "Draft the justification document tonight. Something Marcus can review before I present it to him — I want him to be able to read it and agree it's genuinely his research, not just a cover story."
"I'll have it by midnight."
---
Marcus arrived Monday at 0900. He looked like he'd had a productive week — the quality of someone who'd been working well on something he understood. He set down the folder with the revised consent form before he sat.
"The revision," he said. "I'd like to walk through it with you."
He read through it. The form was cleaner than the previous version — the language about the scope of the practitioner's perception was direct rather than hedged, and the consent question was positioned early rather than buried in the assessment description. The person being assessed would know, from the first paragraph, that the practitioner's class could register patterns outside the explicit scope of the assessment request.
"This is correct," Kael said.
"I sent a version to three assessment practitioners who've been doing similar work — cohort peers I've been in contact with. Two of them are going to revise their forms too." He paused. "The third said the existing form was legally sufficient and didn't need revising."
"What did you say to that."
"I said legal sufficiency and informed consent aren't the same thing." He held Kael's gaze. "They didn't respond."
He looked at Marcus. The man in this room was developing opinions that came from somewhere real. Not rules — actual positions, built from the work. The original timeline's Marcus had never built anything from the question. He'd built his resentment instead.
"There's something else," Kael said.
He put Rowan's draft justification document on the table.
Marcus read it once through without speaking. Then again. Slower.
"The Fenner program," he said.
"The documentation standards in the approval meeting minutes differ from current practice in a specific way." He looked at Marcus. "Whether that discrepancy is deliberate is relevant to a broader investigation you're not involved in. But the research question — why the standards differ — is yours. If you want it."
Marcus looked at the document. "You need someone to pull the meeting minutes from the archive."
"My access route to that archive closed when the black market contact was compromised three months ago." He met Marcus's eyes. "Your practitioner code opens a different door."
"With my name on the log."
"Yes."
A silence.
"Is this research I'm doing for you," Marcus said carefully, "or research I'm doing."
"Both. In that order." He paused. "The justification document Rowan drafted is accurate. The question it describes is a real research question — one that connects directly to your consent framework work. But I won't pretend I'm not asking you to do something that also serves a different purpose."
Marcus sat with this for a moment. "The meeting minutes. What do they document."
"A project approval discussion that included assessment practitioners from the evaluation committee." He paused. "One of those practitioners' participation, documented in an official record, establishes a second-source confirmation for a pattern we've been building evidence around. If the record says what I expect it to say, it substantially strengthens the case."
"The case for what."
"For an external oversight review of the Fenner talent intake program."
Marcus looked at his hands. "I know about the Fenner program. One of the cohort members I've been assessing — she was invited through the financial support footnote. She declined." He paused. "She came to me the next week for a consultation. She was trying to understand what the invitation actually represented."
"What did you tell her."
"That the invitation was structured as support but functioned as recruitment." He looked up. "I told her that because I'd seen the footnote structure in other cases and the pattern was consistent. I didn't have documentation. I had pattern recognition."
He looked at Marcus. "The documentation I need from the archive establishes the pattern isn't accidental."
A long pause.
"I'll pull the minutes," Marcus said.
"Read the justification document first. If you're not comfortable with any part of the framing—"
"I've read it. The research question is mine." He stood up. "I'll file the access request today. The archive staff processes requests within forty-eight hours." He picked up his folder. "I'll send you the document when I have it."
"Send it to Rowan's secure channel. Not to me directly."
"Understood." He moved toward the door. Paused. "The fourth practitioner — the one who didn't respond to my consent form note. I think they're going to have a problem eventually."
"Probably."
"When they do — I'd like to be the one who talks to them first. Before it becomes someone else's problem to fix." He looked at Kael steadily. "Not to protect them. To give them the chance to do it themselves."
He looked at him. "Yes."
Marcus left.
Forty minutes, three separate issues, one of them the archive request. That was what this version of Marcus was — he came with his own problems, his own positions, his own things he was building. Kael was one thread in his week, not the center of it.
He went to circuit work for two hours, then ate, then sat at his desk and opened the training protocol documents he'd been working from.
The C-rank circuit architecture worked on junction density — a process his original training had mapped in detail over months of S-rank practice. He'd remembered, from that original process, a specific methodology for accelerating the D-rank to C-rank transition: a junction resonance pattern that worked by running two channel streams simultaneously at specific interference angles, creating constructive resonance that built density faster than the standard sequential approach.
He'd been building toward trying it. The channel architecture was ready — the junctions were at the threshold density where the technique required.
He ran the circuit work for twenty minutes at standard architecture. Then shifted the approach: brought the secondary channel stream online at the angle the technique required, initiated the interference pattern.
The architecture shut down the secondary stream.
Not dramatically — not a collapse. Just a clean, structural rejection. The junction architecture recognized the interference attempt and responded by routing the secondary stream back to standard. The system had read the pattern and overrode it.
He tried it again. Different angle. Same result.
He tried the third variant from the original methodology. The one that had worked when the first two didn't, that operated through a more complex resonance path.
Same rejection. Cleaner. Like the architecture had already logged the attempt and prepared for the variation.
He stopped.
Sat in the quiet of the training space and looked at the wall.
In the original timeline, that technique had been a workaround — something hunters had figured out through trial and error, passing the methodology between themselves informally because the Association's official training frameworks didn't document it. It had worked because the system's channel architecture had a gap in its optimization logic that allowed the interference pattern to function.
That gap was gone.
Either the system had updated, or the gap had never existed in this timeline, or his changes had created conditions that closed it before he could use it. Three possible causes. The result was the same: the training methodology he'd been planning to implement for the past three weeks didn't work.
The C-rank markers were still a direction. They were just going to need a different path.
---
He didn't tell Elara about the training plan until that evening, when she came over with food from the canal district restaurant two blocks from her building and they sat on his kitchen floor because the table was covered in Rowan's documentation.
"You're going to figure out a different approach," she said.
"Yes." He moved a stack of papers to clear space for the containers. "It's a setback, not an end. The architecture is still at the right threshold — I just need a different way to work with it."
"How long."
"The standard approach would have taken six to eight weeks. This might add two to four weeks to that." He looked at the food. "Castellan hasn't heard about this yet."
"When are you telling her."
"Wednesday session." He paused. "She'll have opinions."
Elara looked at him with the dry quality she brought to things that were obviously true. "She always has opinions."
"Yes." He ate. "She'll probably be right."
She leaned against the kitchen cabinet and watched him. There was something she was thinking about — not hiding it, just deciding whether to say it.
"The archive access," she said finally. "Is it dangerous for Marcus."
He'd told her the framework, not the details. "The log risk is manageable if the committee timeline holds. He knows what he's doing."
"He's doing it because you asked."
"He's doing it because it's also genuinely his research." He met her eyes. "Both things are true."
"I know." She was quiet for a moment. "You're careful about that. Making sure the things you ask people to do are also things that serve them." She looked at the food. "You weren't always."
He looked at her.
She'd said it without edge — an observation rather than an accusation. The way she'd been saying things lately, since the conversation about the regression, where she'd stopped filtering the observations that might land wrong and just said them.
"No," he said.
"You were good at it. But you assumed that what you thought was best was what was best."
"And now I ask." He looked at the window. "It's not generosity. It's accuracy."
"I know." She pushed a container toward him. "Eat more. You're doing training at three thousand meters above your normal altitude and you're eating like it's a rest day."
He ate more. The canal light was doing its thing with the kitchen window — the angle of it in the evenings, the way it moved across the floor in a strip that meant the afternoon was finishing. She sat across from him with her legs folded and her hair down and the quality of someone who was comfortable with this room now, which was different from a month ago.
After the food, the documentation got cleared off the table and ended up in a different pile in the corner, and she stayed.
The canal had gone dark by the time they got to bed, the lights on the opposite bank the only illumination coming through the window. He'd had the thought earlier, during circuit work, that desire was a strange thing to still feel when you were carrying this much — the ethics submission, the archive access, the training setback, ten years of futures you were trying to rewrite. Strange that the body didn't care about any of it. Strange that it just wanted what it wanted.
She moved against him and he stopped thinking about it and paid attention to what was in front of him instead.
She was different from a month ago in this way too — less careful about what she reached for, less polite about asking. She'd learned what she wanted and stopped managing the wanting. Her hands were decisive now in a way they hadn't been, and there was something in the way she moved that had stopped waiting for his cue. He'd noticed it building over the past few weeks and said nothing, because it wasn't the kind of thing you said. You just noticed it and responded.
Afterward she lay against him in the canal-lit dark and her breathing was the specific slow quality of someone who'd arrived somewhere rather than just stopped.
"The second gate notice," she said. "For Nadia."
"Three weeks."
"Is that on schedule."
"Ahead of schedule." He looked at the ceiling. "She doesn't know that yet."
She made a quiet sound that wasn't quite a laugh. "You're going to surprise her again."
"I might." He paused. "She likes accurate information. Accurate information delivered ahead of expected windows makes the information more credible, not less."
"That's not why you're doing it."
He was quiet for a moment. "No."
She didn't push it. She knew why he was doing it — the Blackwood clearance was the real priority, and everything about the Nadia arrangement was calibrated to make that clearance happen on a tighter timeline than originally structured. He didn't need to explain it.
He held her until the canal district's late-night sounds went quiet, and then she went back to her building because she had an early session in the morning.
He lay in the dark and thought about the circuit work.
The exploit was gone. The standard methodology would work, just slower. He needed to talk to Castellan. He needed to understand if there was a legitimate accelerated approach that didn't depend on remembered workarounds.
The training plan was going to have to be rebuilt from different materials.
He went to sleep.
