The procedure started the way procedures did: quietly.
Choi's hands were steady. The cardiac monitor's waveform displayed on Mira's reader — the sixty-four-beat rhythm, regular, the QRS complex morphology within the parameters that the baseline established. The molecular severance tool's disk pressed against the carotid's surface. The calibration complete. The specialist's eyes on the tool's interface display rather than on Shin, the professional's attention allocated to the data rather than the patient.
"Initiating," Choi said.
Shin felt the severance field enter the integration site.
Different from the reinforcement tool's constructive wave. The reinforcement had felt like support, the strengthening of existing structures, the energy supplementing the bonds' resistance. This felt like dissection. The molecular-precision targeting of the interface boundary, the energy not adding to the bonds but teasing at their connections, separating the crystal's molecular layer from the biological tissue matrix at the integration site's hybrid interface.
Not painful. The hybrid tissue's modified sensory properties didn't carry pain signals the way the intact biology did. But the perception's resolution tracked the process clearly: the interface bonds loosening, the crystal-phase material at the boundary beginning to differentiate from the biological carotid layer, the integration front's advance slowing as the severance field disrupted the molecular connections that the advance required.
"Field established," Choi said. "Boundary disruption at eight percent. Integration advance has suspended."
The advance suspended. The one-point-three centimeters per hour had dropped to zero within the severance field's active radius. The integration front was still. The molecular disruption preventing the crystal's advance into the biological tissue.
Mira checked the cardiac monitor. "Heart rate: sixty-four. Rhythm: regular. QRS morphology unchanged. Continue."
The procedure ran.
At fifteen percent: the boundary disruption showing consistent progress. The interface bonds loosening uniformly across the contact area, no asymmetric response, no unexpected tissue reactions. The calibration's accuracy confirmed by the uniform field effect.
At twenty-two percent: the first anomaly.
Not in the cardiac waveform. Not in the tool's interface display. In the perception.
The repair mechanism.
The construct's self-repair mechanism — the process that had been running at two percent per hour following the reinforcement procedure's completion, the crystal's repair energy directed at the sixty-one percent fracture — had reallocated. The energy wasn't going to the fracture anymore. The processing cycle's channel-four allocation had shifted, the repair mechanism's target changing from the internal fracture to the integration site.
The construct was treating the severance field's boundary disruption as damage.
The repair mechanism was running against the severance.
"Pause." Shin said.
Choi's hands stilled. The tool maintained contact — the specialist's training holding the static position that an abrupt withdrawal could destabilize. "Holding at twenty-two percent."
"What do you feel?" Mira asked.
"The construct's repair mechanism has shifted target. It was working on the internal fracture. Now it's working on the integration boundary — the same site the severance is targeting. The repair mechanism is treating the severance field's molecular disruption as structural damage to be corrected."
Choi's expression showed nothing. The professional's processing of the information. "The repair mechanism's response to the severance field. This is a documented possibility in the procedure's risk matrix."
"The risk matrix that I approved." Mira's voice had the precision it got when she was angry and was choosing not to express it directly. "The mitigation listed for this specific risk was 'calibration accuracy reduces the likelihood that the repair mechanism identifies the severance as structural damage.' The calibration is accurate. The repair mechanism identified it anyway."
"The construct's self-repair sensitivity is higher than the standard Resonance Entity profile," Choi said. "Possibly due to the active repair process already in progress on the internal fracture — the repair mechanism is in a heightened activation state, which increases its sensitivity to any molecular-boundary changes."
"Which you would have known about if you'd asked about the internal repair state during setup," Mira said.
A pause. Short. "The active repair was documented in the data I received from Park's procedure assessment. I assessed the heightened sensitivity as within the procedure's tolerance parameters."
"Assessed incorrectly."
Choi said nothing.
Shin looked at the fracture through the perception. The repair mechanism's current target: the integration boundary. The energy previously directed at the internal fracture now directed at the site where the severance field was operating. The repair mechanism's function: reinforcing molecular bonds at the boundary layer, restoring the connections that the severance was disrupting.
The repair mechanism was rebuilding the bond faster than the severance was breaking it.
Not much faster. At twenty-two percent completion — the severance field had disrupted twenty-two percent of the interface boundary — the repair mechanism was running at a lower throughput than the severance. The net effect was still boundary disruption, still progress toward severance. But the efficiency had dropped. The timeline extended.
"How long to complete at this efficiency," he said to Choi.
The specialist ran the calculation. "If the repair mechanism's throughput remains at current levels, the procedure extends from the original forty-minute estimate to approximately seventy-five minutes. The boundary disruption rate is halved by the competing repair input."
Seventy-five minutes versus forty. The colony holding at staging distance. The calcium at seven-point-one, reduced extraction demand from the slowed integration. The cardiac monitor stable.
The operational parameters were still achievable. Slow, but achievable.
"Continue," Shin said.
"Shin." Mira's voice.
He looked at her.
"The repair mechanism is competing with the severance. If the mechanism's throughput increases — if the processing cycle allocates more energy to the repair response as it recognizes the boundary disruption isn't being corrected — the net disruption rate could reverse. We could end up with negative progress. The boundary re-integrating faster than the severance is separating it."
"Current repair throughput trend."
Mira checked the reader. The monitoring interface had been tracking the repair mechanism's energy allocation since Choi's mention of the documented risk. "Increasing. Slowly. Eight percent above baseline when the severance initiated. Currently eleven percent above baseline."
Three percent increase in the time between Shin's first observation of the anomaly and this conversation. The upward trend. Not steep, but present.
"The construct adapts," he said. Not to Mira. To himself, the recognition arriving through the processing of what the perception was showing him. "It adapted to the spider colony's tactics. It adapted to the reinforcement procedure's energy input by activating the repair mechanism. And now it's adapting to the severance by intensifying the repair."
"Adaptive response," Choi confirmed. "Resonance Entities demonstrate behavioral adaptation to environmental threats. The severance field is categorized as a threat."
"How fast does the adaptation accelerate."
"Variable. Entity-dependent. The standard protocol documentation doesn't include data for an Entity in active self-repair mode with an existing integration bond. This configuration is outside the standard profile."
Outside the standard profile. The phrase that had been following Shin since the system classified him at Level Zero.
He sat with the decision for four seconds. The perception tracking the repair mechanism's allocation, the fracture's current state, the integration boundary's disruption progress at twenty-two percent. Mira's reader showing eleven percent above baseline repair throughput and a three-percent increase in the time since identification. Choi holding the tool steady against his neck.
The operational choice: continue and accept the extending timeline with the risk that the adaptation outpaces the severance. Or halt and reassess.
"Continue," he said.
Mira's jaw tightened. One degree.
The procedure resumed.
---
At thirty percent, the pause Mira had requested.
The specialist held static contact. Mira ran the assessment. Three minutes.
The repair mechanism's throughput: sixteen percent above baseline. Eight percentage points higher than at the twenty-two-percent check. The increase rate had accelerated, not linearly but with the curve of a biological adaptive response rather than a mechanical one.
"The net disruption rate," Mira said. Her voice had the flatness she used for numbers she didn't like. "Seven percent. Down from fifteen percent at initiation. If the trend continues at this acceleration—"
"I'll call it," Shin said.
"I know you'll call it. I want you to see the number before you decide to continue past thirty percent." She turned the reader so he could see the trend graph. The repair throughput's upward curve. The net disruption rate's downward curve. The intersection point, the projected crossover where the two lines met and the net progress reversed from positive to negative.
Projected intersection: at approximately fifty-three percent procedure completion.
He'd reach fifty-three percent and then the procedure would start running backward. The repair mechanism overwhelming the severance rate. The interface bonds being restored faster than they were being disrupted.
"At fifty-three percent, continued application produces integration tightening, not loosening," Mira said. "The partial severance creates disrupted boundary zones that the repair mechanism incorporates with different molecular architecture than the original integration — potentially stronger bonding, not weaker."
Choi said nothing. The specialist had seen the same data.
"Halt the procedure," Shin said.
The specialist withdrew the tool from the integration site in the slow retraction that the static-contact holding position had required — the decompression managed to avoid abrupt changes in the boundary's molecular state. The disk cleared the carotid surface.
"At thirty percent completion," Choi confirmed. "Procedure suspended."
The silence in the habitable zone had the specific weight of a room full of people calculating.
Shin tracked the integration site through the perception. The thirty percent of the interface boundary where the severance field had been operating: partially disrupted. The molecular connections loosened, the crystal-biological interface at less than full integration integrity. But the repair mechanism was already responding — the processing cycle's allocation increasing to the integration boundary, the repair energy flowing into the disrupted area.
The partially disrupted area was being incorporated faster than the undisrupted area would have been. The repair mechanism's enhanced-sensitivity response, triggered by the severance field's operation, was now running at a higher throughput than before the procedure began.
"The integration rate," he said to Mira.
She checked. The reading took longer than usual — the repair mechanism's altered energy allocation producing a different signature than the pre-procedure baseline. "Two-point-one centimeters per hour." Her voice was level. "Up from one-point-three before the procedure."
Two-point-one. The procedure had accelerated the integration.
Not by catastrophic amounts. Not the three-point-eight pre-repair rate. But faster than before the severance attempt.
The partial disruption at the interface boundary had made it worse.
Kessler spoke from the two-meter position. The A-rank's voice carried nothing that Shin could categorize as apology. "The interaction between the active repair mechanism and the severance field was a documented risk in the procedure's protocol. The probability was assessed as low given accurate calibration."
"The probability was assessed by people who hadn't examined the specific entity's adaptive parameters," Shin said.
"Correct."
"And you knew the assessment was based on standard-profile entities."
"I knew the procedure was designed for standard-profile entities. The subject is not standard profile." A pause. The A-rank's version of an acknowledgment. "The operation was viable against the probability distribution. The outcome fell in the low-probability range."
The low-probability range. The integration site compromised. The timeline compressed. The severance technique that was supposed to be the solution was now part of the problem.
He'd heard the risk. He'd decided to continue past twenty-two percent when the repair mechanism's increasing throughput had been visible. He'd been watching the trend line. He'd told himself the operational parameters were still achievable.
He'd been wrong.
The mistake wasn't trusting Kessler or Choi. The mistake was his own decision at twenty-two percent, the continuation when the data said to stop. The same calculus error that had gotten him into trouble before: identifying a risk, quantifying it, deciding the operational objective was worth the risk, and being wrong about the calculation. Not wrong about the objective's importance. Wrong about the numbers.
Mira's hand was on his arm. Not in the clinical position. The other kind.
She didn't say anything. The healer who'd told him to halt and been right wasn't interested in being right at this cost.
"The integration site now," Shin said. "What's the state of the partial disruption."
Mira checked the reader. "The partially disrupted boundary zone is being repaired with accelerated molecular incorporation. The repair mechanism's heightened sensitivity state is drawing more processing-cycle energy to the integration site than before the procedure. The fracture repair has essentially stopped — the energy reallocation away from the fracture to the integration boundary has suspended the fracture's self-repair."
The fracture repair suspended. The integration accelerated. Both worse than the pre-procedure state.
"The fracture."
"Sixty-one percent. Still. The repair mechanism wasn't at heightened sensitivity when the fracture repair was running — the two percent per hour was the baseline sensitivity's contribution. At heightened sensitivity, the energy that had been going to the fracture is now going to the integration boundary."
"How long before the heightened sensitivity state normalizes."
"Unknown. I don't have data on Resonance Entity adaptive-response normalization timelines."
Park spoke from the far side of the habitable zone. The Crimson Gate engineer had been silent throughout the severance attempt — the professional's observational mode processing a procedure outside their specialization's territory. "The construct's adaptive response to the reinforcement procedure normalized within six minutes of the reinforcement's completion. The processing cycle's energy reallocation settled back to its baseline distribution after the external stimulus was removed."
Six minutes for the reinforcement's stimulus. The severance's stimulus was stronger; the boundary disruption was a larger-scale intervention than the molecular bond supplementation. But the normalization direction was correct.
"If the heightened state normalizes in fifteen to thirty minutes," Mira said, slowly, working through the clinical model, "the fracture repair mechanism would resume. The integration rate would stabilize at the lower rate — one-point-three or possibly slightly higher. The disrupted boundary zone would continue incorporating at the enhanced rate, but the overall advance would slow back toward the pre-procedure baseline."
Not good. But not irreversible.
Fifteen to thirty minutes of waiting.
Shin looked at the cardiac monitor. The waveform still regular. The calcium still at seven-point-one. The colony still holding at staging distance, thirty-seven meters northeast.
The mistake was made. The consequences were real. The calculus was wrong at twenty-two percent, and the outcome was that he was worse off than before the procedure started.
He needed to sit with that. Not for long. But honestly.
He sat with it.
"Twenty minutes," he said to the room. "We wait twenty minutes for the heightened state to normalize. Then Mira reassesses. Then we figure out what the options actually are."
Vasquez nodded. Hae-won's posture acknowledged. Kessler remained still.
"The meeting," Kessler said. The A-rank's first proactive statement in twenty minutes. "Guild Master Renault anticipated that the severance procedure's outcome would be suboptimal. In that event, he wanted me to say: the offer includes the guild's medical infrastructure. Facilities that have experience with post-integration complications."
Shin looked at the A-rank.
"Fourteen months," he said.
Kessler's expression was the blank professional affect. "The subject's cooperation becomes more valuable as the complications increase. Guild Master Renault understands that position."
The construct pulsed. Sixty-four beats. The integration advancing at two-point-one centimeters per hour through the disrupted boundary's altered molecular landscape.
The instrument panel, reading gauges that were moving in the wrong direction.
He'd made the call. It had been wrong.
Twenty minutes.
