It happened in the third week, on a Tuesday afternoon when a dust particulate filter alarm had been running in the secondary ventilation unit for forty minutes and neither of them had addressed it.

The alarm was minor -- a maintenance indicator, not a system warning -- and it would continue running until someone cleared the filter or manually acknowledged the alert. It was audible in the common area and less audible in the research spaces, which was probably why it had been running for forty minutes without either of them addressing it. Lucan had noticed it approximately thirty minutes ago, had added it to the list of tasks he was going to get to after the current synthesis calculation, and had continued working.

Mirelle appeared at the partition door.

"The alarm," she said.

"I know. Filter maintenance. I'll get to it."

"It's been running for forty minutes."

"I'm aware of the duration."

She looked at him for a moment. "Do you want me to clear it?"

He was about to say that he would do it when he finished the calculation, which was approximately twenty minutes away, and then he looked up from the workstation and found her expression, which was not impatient or irritated -- she did not do either of those things very visibly -- but which had a very slight quality of: I have been listening to this for forty minutes and if you say you'll get to it in twenty more minutes I am going to take the situation into my own hands without further discussion.

He got up and cleared the filter.

The silence, when the alarm stopped, was notable after forty minutes of not-silence.

"Thank you," she said.

"It was my maintenance item."

"You're welcome to it. I was thanking you for doing it before the fifty-minute mark, which I estimated as the threshold of my patience."

He looked at her. She had a dry quality to her speech that he had been observing for three weeks -- a specific flatness in the delivery that contained, on examination, a layer of precision that was not quite humor but was adjacent to it. He had not commented on this before because commenting on it seemed likely to change it.

"Fifty-minute threshold," he said. "Is that a standard figure or specific to filter maintenance alarms?"

She considered. "Probably specific to filter maintenance alarms. The threshold for other categories would be different. I haven't established them precisely."

"That seems like an area worth researching," he said.

She looked at him. The very slight adjustment of expression that was, he had learned, the marker of something landing differently than expected. "You're funnier than you present."

"I have a narrow range of contexts in which it appears."

"Laboratory maintenance failures."

"Apparently."

She looked at the now-silent ventilation unit and then at him. "What other contexts."

He thought about this. "Error correction. When someone is very wrong about something very specific. And occasionally when I've been awake for too long and the filtering stops working properly."

"Which of those applies right now."

"Probably the last one," he said. "I've been awake since four in the morning for the pre-dawn collection run."

She looked at him with the look that meant updating a model. "You do that every day?"

"When atmospheric conditions permit. Approximately four days a week."

"Without telling me."

"It's my collection protocol. I didn't think it required notification."

"I wake up when the hatch cycles," she said. "I've been hearing it and not knowing what it was."

He had not thought about this. He had not thought about the fact that someone in the station who was accustomed to being alert to sounds of access would hear the hatch and have no context for it. "I'm sorry. I should have told you."

"You could leave a note," she said. Not angry -- factual, the way she said most things. "On the kitchen counter. Before you go. Just the time of departure and expected return."

He considered. "That seems like a reasonable system."

"It is. It also means I know where you are, which is separate from whether I'm worried about you, which I'm not, because your surface competence is evident."

He looked at her. "You've assessed my surface competence."

"Yes. Across fourteen days of observation. Your route variation is appropriate, your exposure management is conservative by good margins, and your return timing is consistent. You're not going to die on a collection run." She paused. "The note is still useful for the other reason."

He held her look for a moment. He had been assessing her across fourteen days as well, which he had thought was his own private project, and here was the information that it was not private and had not been. She had been doing the same thing, in the parallel, quiet way she did most things, and she had arrived at results.

"All right," he said. "I'll leave a note."

She nodded. She returned to the analysis room.

He stood in the common area for a moment after she left. He thought about surface competence and about the specific quality of being assessed in return, and about the fact that she had said I'm not worried about you in the same sentence where she said the note is still useful, which was precise in a way that was not comfortable exactly but was something.

He went back to the research wing. He sat at the workstation and returned to the synthesis calculation and found that it was more difficult to concentrate on than it had been before the alarm, which was probably the sleep deprivation she had correctly identified, and possibly something else.

---

In the evening, she came to the research wing again.

She had been coming into the analysis room regularly, but this was the second time she had come to the primary workstation side, and the first time she had come in the evening rather than during the working hours. He looked up when she appeared.

"I've been reading the transit maps," she said. "I have a question about the infrastructure."

"The underground network or the surface routing?"

"Both, at the junction. The Hollowline has a branch line that goes northeast, toward the old industrial district. The maps show it as decommissioned, but the rails are still there on the surface survey."

He knew this line. He had catalogued it in his first surface survey as a potential collection route access point. "It was decommissioned when the surface transit systems were abandoned. The rails are there but the vehicles aren't. The electrical infrastructure along the northeast branch is still partially intact from the last maintenance cycle before decommissioning, which was forty years ago."

She looked at him. Something in her expression was different from the usual assessment -- this was interest, real and present. "The electrical infrastructure. Including the signaling system?"

"The signaling hardware exists. I don't know the current state of the circuits. I've never had a reason to test them." He paused. "Why?"

She was quiet for a moment. "I know people," she said. "In the surface network. People who use the old transit routes. Not officially." She paused again. "I want to know if there's a way to use the old signal system to communicate with the northeast branch network without going through the main settlement infrastructure."

This was the most she had told him about her own network in three weeks. He received it with the care he tried to bring to everything she offered.

"I can run a diagnostic on the signal hardware when I'm in the northeast sector on a collection run," he said. "Without entering the primary network. I can tell you what's functional within two or three runs."

"I'd appreciate that," she said.

He nodded. He wrote it in the collection protocol notes.

She went back to the analysis room. He looked at the note he had written and thought about the networks she moved through, the people she knew in the surface routes, the life she had built in the spaces between official ones. He thought about what it would mean, if the signal system was functional, to have a channel she could use.

He thought about this as a practical consideration and then noticed that he was also thinking about it as something else -- as a way of giving her something that had nothing to do with the research, something that was simply useful to her in her own right. He noticed this and did not look away from it.

He added the diagnostic to his next collection run schedule.

---

The next day he left a note on the kitchen counter before the pre-dawn run.

Surface collection, northeast sector. Departure 05:20, return approximately 07:30. Standard atmospheric conditions. No anomalies on the previous run.

It was more information than she had asked for. He had started with the minimum -- departure time, return time -- and found that he kept adding sentences. He stopped at four lines and went to the hatch.

When he came back at quarter to eight, the note was still on the counter. She had added to it in the margin, in handwriting he recognized as hers: the transit signal hardware is in junction box NE-4 on the map. Small red case, northeast wall.

He read this standing at the counter in his surface kit with the mask still on. Then he took the mask off and read it again.

She had given him specific information about the signal hardware location -- information she would have gotten from her own knowledge of the northeast sector, from the surface network she had been part of, from sources she had not told him about. She had added it to his note. Not a new note, not a separate message. An addition to what he had started.

He left the note on the counter.

He did not annotate it further. He went to shower off the surface contamination and change into his working clothes and the note stayed where it was on the kitchen counter, two sets of handwriting on the same piece of paper.

---

On the twenty-second day, he made a mistake in the synthesis framework.

It was not a catastrophic mistake -- an assumption in the pathway model that he had built into the early stages and that he caught when the third-phase calculations produced a result that was inconsistent with the precursor data. He would need to go back to the second-phase pathway and rebuild it with the corrected assumption, which was approximately two days of work to redo.

He sat at his workstation for a while after identifying the error. Not panicking -- he was not a person who panicked about research setbacks, having had enough of them over sixteen years to understand that they were a normal part of the work rather than a signal of fundamental failure. But sitting with the specific weight of two days lost against the timeline he was working against.

Mirelle appeared at the partition door.

He had not been visibly distressed -- he was not, in general, visibly distressed. But she had been in the analysis room and had apparently noticed something.

"Setback?" she said.

"Correctable one. Approximately two days to rebuild the pathway."

She looked at the workstation display. "Not catastrophic."

"No."

She came into the research wing -- all the way in, to the workstation area -- and looked at the display. He moved so she could see it clearly. She looked at the pathway diagram and the point where the inconsistency had appeared.

"There," she said, pointing to a section she should not, by any training, have been able to identify.

He looked at where she was pointing. It was, in fact, the exact section that was incorrect. "How did you identify that?"

"I've been reading the framework document," she said. "In the evenings. From the server."

He looked at her. She had been reading his research framework document in the evenings, from the station's internal server, without telling him.

"I should have told you," she said, recognizing his expression. "I was going to ask permission and then I did it instead without asking, and then it was done and I didn't know how to address that without making it seem like more of a thing than it was."

He processed this. "How long?"

"Two weeks."

"The whole framework?"

"The whole framework."

He looked at her. He thought about the specific symmetry of this -- she had been reading his research without telling him, and he had found her designation in the database without telling her until later, and neither of them had been doing it to harm the other, and both of them had struggled with how to address it afterward. He thought about the specific quality of two people who were both very careful about information, negotiating with each other in a space where information was shared.

"Thank you for telling me," he said.

"I would have asked first," she said. "I should have asked first."

"The permission was already there," he said. "We discussed this -- I told you I'd share findings as I found them. The framework is a finding." He paused. "I should have shared it explicitly rather than assuming you'd know you could access it."

She held his look for a moment. Then she looked at the display.

"The error is in the binding angle assumption," she said. "You used the standard binding angle from the secondary literature, but the precursor compound from my samples has an unusual steric configuration that changes it. It's in the structural analysis from the third sample."

He looked at the display. He looked at the third sample structural data. He found the steric configuration she was describing.

She was correct.

He sat down. He opened the pathway framework and found the binding angle assumption. He ran the recalculation with the corrected figure.

The inconsistency resolved.

"That's not two days," she said, behind him.

"No," he said. "That's approximately four hours."

He turned around. She was standing at the research wing entrance, at the edge of the workstation area, and she was looking at him with something that was not quite a smile but was the expression that lived in the same space as a smile before it decided whether to arrive.

"You've been reading the framework for two weeks," he said.

"Yes."

"And you identified the steric configuration as a relevant variable."

"It was in the structural data. I noticed it didn't match the binding angle assumption in the framework and I was trying to figure out how to tell you without it sounding like I'd been going through your work without permission."

He looked at her. He thought about the two weeks of evenings she had spent on the framework, and about the correction she had just made, and about the expression that was living in the space before a smile.

"Would you like to see the corrected pathway?" he said.

She came the rest of the way into the research wing.

---

They worked on it together for three hours.

Lucan had not expected this. He had meant show you the corrected pathway as an ending to the scene -- as in: I will display it on the screen and you can look -- and it had instead become something else, a working session, Mirelle pulling the analysis room chair to the side of the workstation and looking at the pathway with the specific quality of someone who was not a trained researcher but who had been, for the past two weeks, a very attentive reader of a research framework and who had, for the past several years before that, developed a working understanding of the relevant biology through direct experience.

She asked good questions. Not always the right questions in the technical sense, but questions from angles he had not considered, perspectives from outside the framework that sometimes identified assumptions he had not known he was making. He answered each one. He also found himself explaining more than he would have explained to a colleague -- more of the underlying logic, more of the historical dead ends, more of why each decision in the framework had been made the way it had been made.

At some point in the second hour, he noticed that he had been talking for twenty minutes and she had been listening with the specific quality of someone who was genuinely following rather than politely attending. He stopped.

"I've been explaining for a long time," he said.

"I know. I've been listening for a long time." She looked at the display. "The pathway is more elegant than it looked in the framework document. Seeing it dynamically changes the sense of it."

He looked at the display. "Elegant" was not a word he usually used about research frameworks. He thought about what it meant that she had used it.

"The synthesis target is here," he said, pointing to the endpoint of the pathway diagram. "If the design holds, and if I can obtain the precursor compound consistently enough to confirm the replication parameters, the synthesis will produce a compound structurally equivalent to what your immune system makes. Which can then be produced without a biological source."

She looked at the endpoint. She was quiet for a moment.

"Without a biological source," she said.

"Yes. The goal of the synthesis is to remove the dependence on a biological source entirely. The compound would be manufacturable through chemical synthesis, which means it could be produced at clinical scale and distributed through the standard treatment network."

She was quiet for another moment. He watched her face and saw the specific, careful quality of someone arriving at the full implication of a thing they had understood intellectually for weeks and were now understanding in a different register.

"If it works," she said, "I become unnecessary."

"From the research perspective, yes." He paused, because precision was important here. "That's the intended outcome of the research. Not to create a dependence on you, but to use what your biology points to in order to build something that doesn't require your biology to sustain."

She looked at the endpoint of the pathway for a long time.

"Good," she said. Quiet. The same word she had used when he confirmed consistent expression. Not warmth, but something that had been through warmth at some point and come out the other side as something more specific.

He did not say anything. The display ran its gentle animation of the pathway, the compound moving through its stages toward the target.

"I want to be there when you confirm it," she said. "Whenever that is. I want to see it."

"All right," he said.

She got up and returned the analysis room chair to its position. She went through the partition, and he heard the map being unfolded, and the research wing settled into its working quiet with both of them in their respective spaces.

He looked at the corrected pathway for a while. He looked at the endpoint. He thought about the word good and about the look on her face when she understood fully what he was working toward, and he thought that this was one of the moments that was going to be in the data set for a long time, one of the ones he would keep returning to.

He turned back to the synthesis design. He had four hours of work he no longer needed to redo, and a steric configuration he had not known about until twenty minutes ago, and a pathway that was cleaner than it had been before she walked through the partition door.