The framework arrived in forty-three hours.
Mira had not slept during most of them. Voss could tell from the quality of the presentation — the same ideas she would have expressed after rest, expressed here with a slight over-precision. The careful articulation of someone compensating for fatigue by adding redundancy to every sentence, as if the work was too important to leave any surface uncovered. He did not mention this.
She walked him through the framework section by section.
"The fundamental problem," she began, "is that we're trying to perform surgery on a living mind. Traditional Thread Severance is designed for ability structures and barrier constructs — things that are complex, but not continuous. They have discrete boundaries. You can identify where the structure starts and where it stops. The severance is performed at the boundary and the removed structure collapses inward."
"Neural pathway architecture doesn't have discrete boundaries."
"Correct. The personality-core threads and the cognitive-function threads are continuous with each other. They're not separate modules with clean interfaces. They're a web — every thread connected to adjacent threads, those threads connected to others, the whole thing running in a dense lattice without natural seams." She sent him the diagram she'd prepared. "Which means Thread Severance in a neural environment can't follow the approach you use for ability structures. You can't find the foreign architecture, locate its primary filament, and cut it loose. In a neural context, there is no primary filament. Every thread you cut takes something adjacent with it."
"Unless you identify the foreign threads individually before severing them."
"Exactly." The pace of her explanation quickened. The point she had been building toward. "Thread by thread. Not cutting the contamination out as a mass. Identifying each foreign filament, isolating its severance point, confirming the surrounding structure is clear, and making the cut with the precision of a procedure rather than an operation."
Voss had been sitting at his desk with the diagram open on his screen. He looked at the highlighted foreign thread clusters. Seven hundred-plus individual threads, by his preliminary mapping. At a conservative pace — the pace that precision required — each one representing a decision.
"The protocol is a mapping phase first," Mira continued. "Before any severance is attempted, we need a complete structural map of the patient's neural architecture in the contaminated regions. Not just the foreign threads. The native architecture too — what's there, where it goes, what each connection does in the larger network. So that when you operate, every cut is planned against a complete picture of what must be preserved."
"How do you map it?"
"Thread Sight, extended and sustained. Not Living Thread Sight — that's limited to seven seconds, which is nowhere near sufficient for a comprehensive map. Standard Thread Sight, reading at depth. You read from outside the patient. You're not inside their cognition — you're reading the structural pattern the way you'd read a building's architecture from outside its walls. Map what you see. Build the diagram."
Voss understood the distinction precisely. Standard Thread Sight let him read composition and structure from the outside, without the immersive Loom-channel connection that Living Thread Sight required. The seven-second cap was Mira's safety measure against the kind of Loom-connection overload that had hospitalized him during the deep Builder communications. Standard Thread Sight didn't carry that risk. He could sustain it indefinitely. Tiring, but sustainable.
"How long to build a complete map?"
"Unknown. We have no precedent." She was honest about the estimate's uncertainty. "The neural pathways are dense. Reading them at the resolution needed to distinguish foreign threads from native threads — to confirm which is which at the individual filament level — requires the finest depth of Thread Sight reading you can achieve. My estimate is six to twelve hours. Possibly longer for Heln, given how advanced her contamination is."
Six to twelve hours of sustained, focused reading. He had held Thread Sight at combat depth for extended periods — he'd read complex Rift structures for hours during clearance operations. That reading was broad and fast, scanning for large-scale patterns. This would be the opposite. Narrow. Slow. Every centimeter documented.
Different kind of effort. He would have to learn it as he did it.
"After the map is complete — the operation itself?"
"You perform Thread Severance on each identified foreign thread using the map as your reference. You know which threads are native and which are foreign. You know the local structure around each target thread. You plan each cut before executing it, verify the path is clear, and make the severance when you're certain."
"The success rate?"
"Theoretically high if the map is accurate and complete." She paused. "The risk is where it's not accurate. Or where the contamination has progressed deeply enough that native and foreign threads are proximate."
"The borderline zones I saw in Heln's preliminary read."
"Yes. I've marked three of them in the diagram. Those are the regions where I expect the discrimination to be most difficult." Her voice was precise. She was telling him exactly what he needed to know, in the order he needed to know it. "In those zones, the margin between a foreign thread and an adjacent native thread is narrow enough that the cut must be very controlled. The angle of severance, the depth, the extension of the cut — all of it matters at a level it doesn't in the clean regions."
"What's the consequence of misjudging in those zones?"
A brief silence. The honest pause.
"Depends on which native thread is nicked. Best case, minor disruption — some temporary fragmentation in a non-critical pathway, the kind of effect that resolves as the native architecture normalizes. A few weeks, maybe less." She paused longer. "Worst case, you sever a core personality thread. The patient survives physically. Their memories are intact. But something central to who they are — some connection that holds the specific configuration of that person together — is gone."
"Permanently."
"Permanently. Thread Severance at that level doesn't heal."
The weight of it settled. Not as abstraction. As three specific people in the medical division — Heln with her flat cataloguing eyes, Grall with his soldier's economy, Ota who had asked whether she would still be herself.
"You're the only one who can do this," Mira said. "The precision Thread Severance requires at this resolution, the ability to discriminate at the individual thread level in a live neural environment — there's no one else. The ability structure for Thread Severance is innate. I can't train someone to it in the time we have. I can't train someone to it in any time."
"I know."
"So I need to know if you're willing to proceed with these parameters."
He looked at the diagram. The borderline zones highlighted. The hundreds of threads that needed to be removed from the architecture of a person who was still very much a person, whose continuation as herself depended on his ability to distinguish one filament from another in a space where the difference between the two was the width of his attention.
The answer had been determined before she asked it. The question was whether he was choosing it consciously rather than defaulting into it.
He was choosing it.
"Yes," he said. "Walk me through the notation system."
---
They spent the next eight hours building the protocol.
Mira adapted the three-dimensional notation grid she had developed for Weaver communication analysis — the same vocabulary of thread types, frequencies, and compositional markers, refined for the finer distinctions required here. She extended it with notation for native versus foreign thread identification: the periodic structure of Memory Thread residue, the organic variability of human neural thread, the frequency signatures that would allow him to distinguish them in the borderline zones when the structural appearance was ambiguous.
Frequency was the most reliable discriminator. Human neural threads operated at a range of frequencies determined by individual cognitive architecture — varied, personal, the biological equivalent of a fingerprint. Memory Thread residue had a characteristic frequency that fell outside that range. In clean regions, the difference was obvious. In borderline zones, the foreign threads had adopted some of the surrounding architecture's variability. The frequency difference was still there. It was just smaller.
He would need to read at depth to find it. But it would be there.
He practiced on neutral material first. A length of barrier-treated cable, its thread composition mapped in the notation system at the finest resolution Thread Sight could achieve. Mira observed through a secondary monitoring feed and confirmed his notation against the known specifications. He could do it. The pace was slow. The attention it required excluded everything peripheral — the sounds of the building, the light changing through the window, the awareness of time passing. Only the reading existed.
That was the right state. He would need to sustain it for twelve hours.
Toward the end of the practice session, he attempted something more difficult: mapping a region of material where two different thread types ran close together, simulating what the borderline zones in Heln's architecture would require. The cable section contained a junction point where the barrier-treated threads met the standard structural threads — different compositions, different frequencies, running adjacent. He mapped the junction at maximum depth resolution.
He could discriminate them. The margin was narrow and the pace required to be careful was slow — much slower than his initial pass through the cleaner regions. But the discrimination was consistent. Thread by thread, he could tell what was what.
The question was whether he could sustain that level of discrimination for the full twelve hours of the mapping phase and then for the hours of the operation afterward. That was a question about endurance rather than capability. He would find out.
At the end of the eight-hour preparation, they had a protocol. Imperfect. The kind of document that would be revised the moment it encountered the reality it was designed for. But complete enough to begin.
"There's one more consideration," Mira said, before he left.
"Say it."
"Your own thread architecture. The Loom-connection stress from the deep communication sessions. You've been accumulating damage in the neural pathway structures that interface with Thread Sight." She was careful, not alarmist. "The mapping phase and the operation are going to draw heavily on those same structures. I need you to be honest with me about how you're feeling during the process. If you notice degradation in reading depth — if the discrimination starts to blur — you stop. We don't push through."
"If I stop, Heln's procedure is incomplete."
"If you push through degraded perception and nick a personality-core thread, her procedure is worse than incomplete." She paused. "You know this."
He knew it. "Agreed."
"If the reading degrades, you stop and we reschedule for the next day. The contamination timeline gives us room for one delay. Not two."
One delay. That was the margin.
He left the preparation session and found Ryn in the corridor. She had been working nearby, not waiting for him — she had her own intelligence work, her own files. But she had registered that he was in the building and had found reasons to stay close to the floor he was working on.
She looked at his face when he came out. Read it the way she read everything.
"It's possible," she said.
"Yes."
"And it's risky."
"Yes."
She walked with him toward the building's exit. Outside, the late evening was clear and cold. The city's lights reflected off low cloud.
"Do you need anything before tomorrow?" she asked.
"No."
"Then I'll make food and you'll sleep."
It was not a suggestion. He did not treat it as one.
"When do you want to start the map?" Mira asked.
"Tomorrow morning. After I've reviewed the protocol one more time and slept." He paused. "I need to speak to Heln first."
"She needs to consent."
"Yes."
He went to the medical division. Heln was awake. He sat down and laid out the protocol — mapping phase, operation, risks. The borderline zones. The specific possibility of cognitive disruption if a cut ran closer than intended. The worst case, stated plainly, because she had earned the full architecture of the decision she was being asked to make.
She listened without interrupting. When he finished, the silence lasted a moment.
"You're asking because there's a real chance it goes wrong," she said.
"Yes."
"And if I don't consent, the alternative is the contamination completes."
"Yes."
Another silence. Shorter.
"Then let's get the map made," Heln said.
The medical bay was quiet around them. The monitoring equipment ran its checks. The overhead light was steady and white.
Outside the window, the city was still lit.
