What the Diagnostic Shows

Sonn arrived at 0600 with a data case that had been repaired twice along one corner seam and a travel mug of something that steamed in the hub's recycled air. They set the case down on the relay table and looked at the state of the hub with the careful, unhurried attention of someone who had learned to read a room from its physical evidence: Denavi at the equipment ledge with her after-action notes, Rovan at the relay panel with one arm in a partial medical brace, the general arrangement of people who had been in the same space for too long and had stopped noticing each other the way you stop noticing furniture.

"The written analysis is complete," Sonn said. "But I want to run the diagnostic in person before you read it."

I was already up. I had slept four hours and forty minutes on the hub floor, which was what I was running on now, and the extension had been processing in the background the whole time and I was functional.

"Here?" I said.

"Here is fine." They opened the data case and retrieved the diagnostic array: the portable sensor web and the interface junction reader and the third piece, the calibration stone they had used since the first diagnostic on the station, now worn smooth along one edge where they always held it during readings. "I want environmental context. You have been in active load conditions for seventeen hours and you have been at reduced load in the same physical space since the engagement ended. The readings will include that context and that is useful."

I sat down at the relay table across from them and held out my left arm with the interface junction exposed. The junction itself had not changed visibly since integration: the same flush-mounted contact points along the inner forearm, warm to the touch, the surface doing the thing metal did not do when you pressed it. But the readings coming out of it were not the same as they had been before the assault, and both Sonn and I knew that.

Sonn attached the sensor web with the precise, unhurried movements of someone who had done this enough times that the procedure was below the level of conscious attention. They pressed the calibration stone against the primary junction point and held it there.

The diagnostic state settled over me the way it always did: a deepening of the extension's background hum, the system's self-reporting coming into foreground clarity. The tone in my right ear shifted up a half-frequency in the characteristic way that meant the interface was under direct assessment rather than passive running. It was not unpleasant. It was simply there.

Sonn watched the readings for three minutes without speaking.

"The stabilization is complete," they said, not to me but to the data pad where the readings were translating into notation. "The predictive extension's new integration threshold has fully embedded. It is not provisional structure. It is load-bearing."

"You said that last night."

"I said it last night based on preliminary data. I am saying it now based on the overnight readings confirming that the threshold did not retract during low-activity rest phase." They adjusted the calibration stone slightly and read the response. "This matters because in two of the historical cases where a new threshold emerged under operational stress, the structure appeared stable immediately after the engagement and then partially retracted within twelve to sixteen hours. You are outside that window."

"And the partial retraction cases," I said. "What caused it?"

"The integration attempted to build structure it did not have the biological substrate to support permanently. The structure held under load because the system was pulling from emergency reserves, and when the load ended, the reserves returned to baseline and the structure could not hold its own weight." They entered a notation. "You do not have that problem. The overnight readings show the new threshold drawing from permanent circuit rather than reserve capacity. The system built it correctly."

I had not known this was a risk until it was apparently not one. I filed that away.

"The visual acuity in low light," I said. "Still at the new level?"

"Yes. That is a permanent structural change, not an enhancement state." Sonn entered another notation. "Your right-ear baseline frequency has shifted, as you know. The low-light pathway adjustment is the other permanent physical modification from this engagement cycle. The two are related: both are sensory-processing expansions that the system built to support the predictive extension's data intake requirements. Your nervous system needed more input bandwidth and the integration created it."

Denavi had gone still at the equipment ledge. She was not looking over, but she had stopped writing.

"What does the development curve look like from here," I said, "given the new baseline?"

Sonn considered this with the pause they used when the honest answer was going to be less complete than they wanted it to be. "The curve has steepened significantly since operational deployment. Each engagement cycle has produced more development than the preceding isolation period produced in the same elapsed time. The system learns more under real load than under training conditions." They looked at the calibration stone's contact reading. "At the current rate of development, your integration state in six months will be meaningfully beyond what I can model based on existing case data."

"Meaningfully beyond," I said.

"I do not have reference points. The Thuvari cases stopped developing before they reached the predictive extension threshold. The Dominion cases are not in the Thuvari records in any detail. You are the only person in the Thuvari database who has developed this particular capability, which means my models are extrapolations from adjacent data rather than projections from established trajectories." They set the calibration stone down on the table. "I am telling you that I cannot tell you with confidence what you will be in six months. I can tell you what you are now, which I can describe precisely."

"Tell me what I am now."

Sonn pulled the data pad and read from the notation. Not for the pad's benefit. The notation was already entered. This was the precision ritual they used when the information mattered and they wanted to deliver it without approximation.

"Primary integration: fully embedded, running at permanent load, zero measurable risk of cascade failure at current development state. Predictive extension: stabilized at the new threshold, integrated into the primary circuit rather than operating as a separate function, developing further in the background based on stored engagement data from the seventeen-hour operational period. Pattern recognition: operating continuously at six simultaneous positional variables in rest state, capable of expanding under active load to a range that exceeded six during the assault but that I cannot define precisely because my monitoring was remote during the highest-load moments." They turned the data pad face-down. "Physical modifications: permanent. The right-ear tone, the low-light pathway, the sleep architecture. You will continue to sleep four to five hours and feel functional on it. This is not a deficit state. It is the new baseline."

I waited.

"There is one more reading," Sonn said. "The stored engagement model of Eshet's integration signature."

"What does the reading show about it."

"The extension built a detailed model from seventeen hours of proximity and engagement. The model is more complete than I expected based on the reported interference from her signal density." Sonn looked at the data pad without turning it back over. "It is complete enough that my working estimate is that you would now identify her integration signature at a range of two hundred to two hundred and fifty meters in a low-interference environment. Against open terrain. More in enclosed structure where the signal reflects."

I sat with that number for a moment. Before the assault, my maximum range for a dense Aeluun signature had been roughly fifty meters based on the probing contacts. The seventeen hours of working against her signal at close range had built a reference architecture the extension was using to identify the specific pattern rather than reading raw signal density.

"The model will update against real-time data if there is a future engagement," I said.

"Yes. The model will refine every time you encounter her. Each contact makes the next identification more accurate." Sonn folded their hands on the table. "I want to ask you something."

"Go ahead."

"The interference you reported during the assault, the moments when her signal density was close enough to disrupt your predictive extension rather than simply be difficult to read. Did the disruption feel like a ceiling you could not exceed, or did it feel like a capacity problem that more development would resolve?"

It was a good question. I thought about the worst moments in the third hour of the assault, Eshet's gifted soldiers moving through the north breach at angles the predictive extension was producing two seconds late on rather than the four-second advance it was giving me for the ungifted soldiers. The gap in the Greel situation. The moment the gifted soldier got through the interior line.

"Both," I said. "At the worst moments it felt like a ceiling. But the ceiling moved over the course of the engagement. The early hours were worse than the later hours, and the reading I had of her timing in the final exchange was clearer than anything I got in the first phase of the assault."

"The model was building while you were using it," Sonn said. "The extension was improving its read on her signal in real time."

"Yes."

"Then it is probably a development problem rather than a fundamental incompatibility." They said it neutrally, the way they said things that had implications they were choosing not to foreground. "If you engage her again at a comparable proximity and the model has had time to process the stored data, the interference will be reduced."

Rovan had been pointedly focused on the relay panel for the past ten minutes. Denavi had gone back to her notes but her pen had not moved.

"The cascade failure risk," I said. "You flagged it in Chapter 10 because of the Thuvari historical cases. What does the current reading say about it."

"The cascade failure cases both involved a rapid escalation of integration load without sufficient biological adaptation between escalation events." Sonn retrieved the calibration stone and turned it over in their hands, the motion they used when they were thinking through something carefully. "You have experienced significant escalation events: the operational deployment, the predictive extension emergence, the sustained seventeen-hour load. But between each escalation, the system has adapted before the next one occurred. The adaptation is not planned or scheduled by any external protocol. The system does itself, building the biological substrate for the next threshold before it is needed." They set the calibration stone back in the data case. "The cascade risk is significantly lower than I believed it was in Chapter 10 because I did not fully understand how the integration managed its own escalation rate. It appears to have a self-limiting mechanism that the Thuvari historical cases did not trigger because those cases involved agricultural and architectural use rather than sustained high-intensity combat."

"Meaning the combat is what activated the self-limiting mechanism."

"Meaning the intensity of the stress load is what activated it. The system built the safety architecture because the load required it. Under lower-intensity use, it apparently does not build that architecture and the escalation, when it comes, can outpace the biological substrate." They looked directly at me. "You are safer than a Thuvari farmer who pushed their integration past normal operational parameters, not because you are stronger or better adapted, but because the sustained load forced the system to build the structural support before the next threshold rather than after."

"That's an odd thing for a safety system to do."

"The Aeluun were not primarily concerned with safety in the way the Thuvari were when they studied the technology. The Aeluun built something optimized for high-intensity use. The Thuvari attempted to use it in low-intensity conditions and found it was not optimized for that, and attributed the failure mode to the technology rather than to the mismatch between the technology's design parameters and their use case." Sonn closed the data case. "I have been thinking about this for two weeks."

"What changed your thinking."

"Watching the system's behavior under your operational conditions. The self-limiting mechanism, the adaptation rate, the way the predictive extension emerged during a high-intensity engagement rather than degrading your function during it. The Aeluun built something designed to work under pressure. The design assumptions were correct, the Thuvari testing conditions were not representative, and the conclusions the Thuvari drew about the technology's danger were partially accurate for agricultural use and significantly less accurate for what you are doing."

Denavi's pen was moving again. I did not know whether she was documenting her after-action notes or something else.

"Write that into the analysis," I said.

"I already have." Sonn looked at me with the directness they used for statements that mattered. "Derek. The reading says you are not in the danger zone I described in the first months. You are in territory with no existing map and some of the unknown variables carry real risk. But the specific risk profile I outlined, cascade failure from rapid escalation, is not where you are."

I heard the thing they were not quite saying: that they had been carrying the weight of that original risk assessment across every engagement, every diagnostic, every conversation, and the current reading was releasing some of that weight.

"I know," I said.

Sonn was quiet for a moment. Then they said: "Whatever the Thuvari were afraid of when they put this technology away, I do not think it was you."

I considered that for a few seconds. "Give it time."

Sonn laughed. It was short and genuine, the laugh of someone who had not been expecting it and did not have the professional composure in place for it. Denavi looked up from her notes. Rovan turned around at the relay panel.

I had not heard Sonn laugh before.

The sound settled into the hub's ambient noise and was gone, and Sonn straightened slightly in the way they did when they had been more open than they intended and were returning to their working register. "The written analysis is on the data pad in the case. The sections on the third candidate's projected development rate are near the end. They are flagged."

"Denavi told me to have the specific ask ready before I talk to Vasht about the contact protocol."

Sonn glanced at Denavi. "Denavi is correct."

"I know," I said.

I pulled the data case toward me and retrieved the data pad. The analysis was forty-one pages of notation in Sonn's precise documentation style, the kind of writing that never reached for effect and hit every measurement with specificity. I went to the flagged sections at the end.

The third candidate's profile was there in the specific numeric language of the integration assessment: compatibility score, biological marker distribution, neural pathway density in the relevant sensory regions. Sonn's projection for development rate under operational stress was highlighted in the assessment field and annotated with a single line that said: Based on the Derek Caulfield development case as reference. No other applicable case exists in the Thuvari record.

The projected rate was two point three times my observed rate in the first operational month.

I read that number twice.

"Nineteen years old," I said.

"The biological substrate is more plastic at nineteen. The system's adaptation rate scales with neural plasticity within certain parameters." Sonn had the data case closed now and was sitting with their hands folded on the table. "The projected rate assumes similar operational stress conditions to the ones that drove your development. In protected research conditions, the rate would be lower."

"And if the Dominion reaches this candidate before the Compact does."

"Then the Dominion determines the operational stress conditions," Sonn said. "And the operational stress conditions the Dominion applies to a newly integrated subject are not gentle."

The hub core hummed. The tone in my right ear ran at the new baseline, the permanent marker running at the frequency it would run at for the rest of my life.

I thought about the first integration, the three hours of calibration that Sonn had described in careful neutral terms and that had felt like my nervous system running every stress response it had ever built in simultaneous replay. I thought about what that felt like at twenty-two with a framework for what it was and eighteen months of context for why I had agreed to it.

Then I tried to think about what it would feel like at nineteen with no framework and no context and a Dominion officer explaining what it was after the fact rather than before.

I stopped trying to think about it.

"The specific ask," I said. "First forty-eight hours post-extraction, in person, direct integration monitoring authority, specific to the third candidate. Technical basis is your analysis of the development rate projection and the real-time monitoring gap between remote Thuvari assessment and on-site integration observation."

"That is accurate," Sonn said.

"And I want you on the relay during those forty-eight hours. Not remote monitoring and a twelve-minute response gap. You, on a live relay, with access to the junction readings in real time."

Sonn's expression did not change exactly, but something in the set of their shoulders shifted. "Vasht will need to allocate secure relay capacity."

"Vasht can allocate secure relay capacity."

"Yes," Sonn said. "She can."

"Then that is part of the ask."

I entered the specific ask into the message draft I had started after Denavi's observation last night and read it back against what Sonn had given me from the analysis. The framing was narrow enough to be a specific request rather than a general claim on the contact operation. The technical basis was clear and linked to Sonn's notation. The relay requirement was stated as a monitoring necessity rather than a personal preference.

Denavi had been right. Framed as coordination rather than a demand, it was harder to decline without an operational reason.

I sent it to Vasht.

Sonn was packing the diagnostic array into the data case with the same precise movements they used for everything. The calibration stone went in last, settled into the case's fitted interior where it had worn a slight impression in the lining material over time.

"The candidates," I said. "When Vasht's contact team initiates extraction, I need to know before the extraction is complete. Not after. Before, with enough lead time that I can be in position for the forty-eight-hour window to start from the moment they are off-planet rather than from the moment they arrive at wherever the Compact stages them."

"I will note that in my own communication to Vasht," Sonn said. "It is a monitoring requirement, not a preference."

"Yes."

They clicked the data case closed. "The analysis is yours. Read the sections on the physical modification timeline when you have time. Not the urgent sections. The ones at the end, past the flagged projections." They held my gaze for a moment. "They are worth reading."

I put the data pad in my cargo pocket.

Sonn stood and picked up the data case and the travel mug, which had gone cold. They moved toward the hub's north access and stopped in the frame with the characteristic pause they used when they had a final thought they were deciding how to deliver.

"The development curve," they said. "The unknown trajectory. The fact that I cannot tell you what you will be in six months." They turned the data case handle in their hand once. "I want you to know that I am not telling you this because I am concerned about the outcome. I am telling you because I believe you should have accurate information and the accurate information is that the map ends where your case history begins."

"I know."

"Yes," Sonn said. "But knowing it and sitting in it while the system continues developing in ways I can measure but not fully interpret are different conditions." They looked at the case in their hand. "I wanted to name that difference directly rather than leave it embedded in the analysis notation."

This was Sonn's version of saying something personal. I recognized it because it was structured the same way Durrik's stories were structured: information delivered as information, with the thing underneath available for whoever wanted to receive it.

"I read the notation," I said. "I will read the sections at the end."

Sonn nodded once and went out through the north access.

Denavi looked up. "The third candidate." She had stopped the pretense of working on her notes. "You're going to be the one who explains what the integration actually is before the decision is made."

"Yes."

"Good." She went back to her notes. "Someone should."

Rovan turned from the relay panel. His arm was in the partial brace but his movement was easy, the discomfort managed down to something he was working around without mentioning it. "The analysis," he said. "Sonn's projection on your development rate. It is public to the unit briefing chain?"

"No."

"Should it be?"

I considered that. The unit had operated with partial information on my integration state across the entire engagement, which was standard practice in Compact operational protocol: need-to-know, and the soldiers needed to know capability and limitations rather than mechanism. But the unit had also spent seventeen hours of the assault relying on that capability and adjusting to its limits in real time, and the people who adjusted to it most effectively were the ones who had developed an intuitive working model of what it could and could not do.

Denavi had that model. She had built it from observation and from the practical experience of the terrain-and-movement partnership.

"The capability summary," I said. "The range, the load limits, the degradation indicators under sustained use. That goes to unit briefing. The mechanism and the development projections stay in Sonn's analysis chain."

Rovan nodded. "That works."

The secure link chimed. Vasht, responding to the specific ask. The response was direct and shorter than I expected: Authorized. Relay capacity allocated. Sonn notified. Contact window for the third candidate is your operational responsibility for the first forty-eight hours post-extraction. First and second candidates follow standard protocol with Sonn remote monitoring only. I am sending the complete extraction timeline in the next communication.

I read it twice to make sure I had not missed a condition I had not agreed to.

I had not.

The specific ask, framed as coordination rather than a demand. Denavi had been right.

The extraction timeline arrived four minutes later: the Compact contact team was already staged, the initial survey of all three candidates was complete, and the operational window for extraction initiation was ten days out. The third candidate was second in the sequence, not last. The sequencing was based on geographic access rather than development priority.

Second in the sequence meant the first candidate's extraction would be complete before the team reached the third. That meant a potential twelve-to-fourteen day gap between the first extraction's completion and my window with the third candidate, during which the third candidate was still on Earth with a Dominion contact operation moving toward them at the same timeline Vasht had estimated.

I drafted a second message. The third candidate's development rate projection made the sequencing a priority consideration. If the Dominion contact was moving at the estimated timeline, the geographic access rationale for second-sequence positioning needed to be weighed against the development rate risk. Requesting extraction sequence review.

I sent it.

Rovan watched me work through this without comment, which was one of the things I had noted about him in the three weeks since he had joined the unit: he tracked operational situations carefully and he did not fill silence with noise. It was a different quality from Hetch's reflexive energy, and it was useful.

"The relay panel is showing a new signal," he said. "Dominion surveillance, passive range, about three kilometers east. It has been sitting there for forty minutes."

I pulled the extension's monitoring output. The signal was there in the passive-range data, low-intensity, the type of persistent positioning signal that Eshet had been running before the assault to map the defensive layout.

"She's already collecting new data," I said.

"Yes," Rovan said.

"Log it. Send it to Greel's medical team relay and to Vasht. Surveillance posture, not active approach, but the position is closer than the pre-assault surveillance baseline by about eight hundred meters."

Rovan turned back to the panel and entered the log notation.

Denavi did not look up from her notes. "She knows you can identify her signature now. She is not going to walk the surveillance asset into your range."

"No," I said. "She is going to walk it close enough to confirm I am still here and still running the extension."

"Which you are."

"Which I am."

The tone in my right ear ran at the new baseline. The extension's passive monitoring held the surveillance signal at three kilometers east, patient and low-intensity, a Dominion Blade-Commander conducting reconnaissance with the methodical care of someone who had been in this work for eleven years and had never needed to hurry an assessment.

I went back to the analysis sections Sonn had flagged and started reading. The technical notation was dense in the way Sonn's work was always dense: every measurement cited, every projection's basis explained, every place where the data ran out clearly marked rather than papered over with approximation. I appreciated that. I always had.

The sections at the end, past the projections, were different. Less notation, more narrative. Sonn had written them in their natural voice rather than the formal assessment register, and the shift was visible in the first sentence of the section.

The section was titled: On the Question of Continuation.

I read it through once without stopping, then went back and read the first three paragraphs again.

What Sonn had written was this: that the Aeluun system, based on everything observable in the Derek Caulfield case history, appeared to have found what it was designed to find. Not a soldier, not a research subject, not a test case for the Compact's military development program. A person whose specific problem-solving architecture, the terrain-and-patience framework that Durrik had identified in the live-fire simulation, the pattern recognition that had emerged from years of reading environments for information rather than for safety, the instinct for exit routes and alternative approaches and the tactical patience to wait for the right moment rather than the first moment, was what the system had been optimized to amplify.

The Aeluun had not designed the technology for agricultural use or for construction or for war. They had designed it for whoever the system determined could make the best use of it. The Thuvari had put it to agricultural use because that was what Thuvari needed. The Dominion had put it to military use because that was what the Vorath needed. The system had responded to both, had built what each user needed from what each user already had.

What Sonn could not determine, and said so directly, was whether the Aeluun had anticipated what the system would do with a human.

I sat with that for a moment.

The last paragraph of the section was a single notation in Sonn's formal notation style, which was the choice of someone who had been writing in their own voice and had needed to step back into the professional register for the final line, because the final line needed to be accurate rather than personal.

Current assessment: subject stable, developing, no mapped endpoint. Recommend continued operational integration pending Earth contact operation. All known risk parameters within manageable range. Unknown risk parameters cannot be assessed from existing case data. No recommendation is possible regarding long-term outcome. Continue gathering data.

I closed the analysis.

The hub core hummed. The surveillance signal sat at three kilometers east and did not move. Vasht's response to the sequencing request arrived: Under review. Will communicate within two hours.

Denavi finished her notes and set the data pad down on the equipment ledge.

"Well?" she said.

"Stable," I said. "Developing. No map past the current position."

She considered this. "But stable."

"Sonn's word."

"Good." She picked up her kit and stood. "I am going to get four hours. Greel's debrief is at 0900 and I want to be rested for it."

"Sleep."

She went out through the north access.

Rovan logged the surveillance signal's position update, because it had drifted fifty meters south in the last ten minutes, still passive, still not approaching, but not stationary either. I watched the drift in the passive monitoring output and let the extension track it without trying to predict where it was going, because at three kilometers and low intensity the predictive model was working from too little data to generate reliable forward projections.

The system knew its own limits.

So did I, more or less. More than I had eighteen months ago on a ridge in Cocke County with a hand-fitted rifle and a deer I was never going to get back to. Less than I was going to need by the end of the ten-day window before the extraction team moved on the first candidate.

I pulled up the candidates' files and read the third candidate's profile again, not looking for new information, just running it through the current state of the integration's working memory. The nineteen-year-old in the southern hemisphere with the compatibility score that had made the Thuvari assessor choose their words carefully. The development rate projection. The fact that the Dominion contact operation was moving at the same timeline.

The tone in my right ear ran steady.

The system ran in the background with the patient thoroughness it always had.

I had work to do before 0900.