The restaurant was on a side street off Newington Road, small enough that eight people filled it comfortably and large enough that the owners had learned not to mind when a particular table stayed occupied for three hours on the last Friday of the month.

Sicheng had been coming here long enough that the owner's daughter, who managed the front of house, brought his usual order without asking. He sat near the window, in the seat he always took when he arrived early, and watched the rest of the group filter in over the following fifteen minutes: Wei-Lin first, always punctual, engineering PhD, the person in Edinburgh he had known longest and trusted most completely. Then Daniel and Marcus — the Marcus who was unrelated to the Marcus in the Whitmore lab, a fact that had required clarification once and that Sicheng found inexplicably funny. Then four others in various configurations of arrival, shedding coats, switching between Mandarin and Taiwanese Hokkien and occasionally English in the fluid way that happened when a group had been talking to each other long enough that the language of any given sentence was determined by what fit it best.

This was the only place in Edinburgh where Sicheng did not perform anything.

It was not that he performed in other contexts exactly — he was not dishonest, did not maintain a false persona, was recognizably himself in seminars and lab meetings and Thursday methodology discussions and hallway exchanges and shared documents at 6:47pm. But there was a register of himself that existed only in Mandarin and Hokkien, only with people who had known him before Edinburgh had shaped him into whatever Edinburgh had shaped him into, only at tables like this one where the food was correct and the conversation moved without requiring anyone to translate themselves.

He had needed this, in year one, more than he had expected to. He had thought, coming to Edinburgh, that he was sufficiently equipped for the specific labor of being foreign: the translation not just of language but of self, the daily work of being legible to people who had different baseline assumptions about the world. He had been wrong in the way you were wrong about something you had not yet experienced — accurately estimating the task but failing to account for the accumulated weight of it, the particular tiredness that built up when every interaction required an additional layer of attention to register and context and presentation.

This table had been, in year one, the place where he set that tiredness down.

He ate his soup. He listened to Wei-Lin describe a problem with her simulation that had been frustrating her for two weeks. He contributed to a conversation about a mutual friend who had just accepted a postdoc in Boston and whether Boston was livable. He was, by all available measures, present and comfortable and entirely himself.

---

"I heard you're collaborating with the Whitmore lab," Wei-Lin said, at the point in the evening when the main dishes had arrived and the conversation had settled into the comfortable sprawl of people who had been talking for an hour and could now talk about anything. She said it the way she said most things — directly, without particular inflection, as a request for information rather than a comment on anything. "The mainland Chinese girl."

The table reconfigured slightly in the way tables reconfigured when a statement landed that had some weight to it. Not hostile — just attentive.

"Collaboration," Sicheng said. "Methodology alignment. She's in her first year with the Whitmore group. Her approach to dynamic boundary conditions addresses a limitation in the Henderson framework." He said this in the same register he used in seminars — accurate, sufficient, aimed at the informational content rather than anything else.

He was aware, even as he said it, that he was monitoring the register. Choosing, at some level he had not decided to engage, what to include and what to leave in the category of things that were accurate but not required. He had not decided to do this. He noticed that he was doing it. He added this to the list of things he was noticing about himself in Edinburgh, which had become, over the past eleven weeks, a longer list than he had expected to compile.

"How long?" Wei-Lin asked.

"Eleven weeks."

She nodded and took a piece of something from the communal dish and said nothing further, which was how Wei-Lin indicated she had received sufficient information for the moment. Sicheng returned to his soup.

---

Daniel, who had been in Edinburgh for six months and had the particular observational intensity of someone still building their mental map of a new city, said: "Taiwanese people and mainlanders don't really mix here. Have you noticed? Different background, different context. Even in the international student spaces, people end up separating."

He said it as an observation. Sociological, dispassionate, the way you described a phenomenon you had noticed and found interesting. Several people at the table made sounds of agreement or recognition.

Sicheng looked at his glass.

He thought about the canteen. He thought about Oliver saying *they've got a thing with mainlanders* in the same dispassionate observational register, as a fact about behavior rather than a judgment on it. He thought about the way she had set down her fork and looked up and said *that's not what it is* in a room full of people who were going to remember she said it. He thought about crossing the seminar room at a particular pace, about the specific mechanics of placing a paper's name in a conversation so that dismissing the paper required engaging with its content.

He thought about the fact that he was sitting at a table of Taiwanese people in Edinburgh thinking about a mainland Chinese person, and that this was not the first time this particular irony had occurred to him, and that it had been occurring to him with increasing frequency since approximately the third week of the collaboration.

He thought about the language switch in the conference room — how it had not been decided, how it had happened because the argument was moving faster than either of them could manage in a second language, how *不介意* had come out of him before he had determined whether he minded or not and had turned out to be accurate. He thought about how she argued: not to win but to locate the true position, and how he had been, over eleven weeks, finding this increasingly difficult to maintain a professional distance from.

He did not say any of this.

"It's different when you're working with someone," he said. "The research makes a different kind of context."

Daniel considered this and nodded. The conversation moved on.

---

"Sicheng."

Wei-Lin said his name the way she said it when she wanted him to pay attention to something — not loudly, not urgently, just with the particular precision of someone who had known him long enough to know that his name in that tone meant *I'm about to ask you something real.*

He looked up from his glass.

"Is she good?" Wei-Lin asked. "In the lab."

Not: *is the collaboration productive.* Not: *is the methodology working.* Is she good, in the lab — which was how Wei-Lin asked about people when she wanted to know about people rather than about their outputs.

He looked at Wei-Lin across the table. Wei-Lin had the patient expression of someone who had already formed a hypothesis and was giving him the opportunity to either confirm it or provide contradicting data.

"She's very good," he said. He said it without any of the careful register management he had been applying to the rest of the conversation. It came out simply, the way accurate things came out when you stopped thinking about the context and just answered the question.

Wei-Lin held his gaze for approximately two seconds longer than the answer required. Then she nodded and reached for the communal dish and said: "Good," and the conversation moved to something else.

He understood that Wei-Lin had understood something. He understood that she would not raise it again tonight, and possibly not ever, because that was how Wei-Lin operated: she noted things, she filed them, she did not require them to be discussed unless they needed to be.

He wondered, not for the first time, whether this was a trait he had developed independently or whether he had learned it from her.

---

The group dispersed in stages — some to a bar further down the street, some directly home, the usual Friday-night calculus of energy and obligation. Sicheng and Wei-Lin left together because they walked in the same direction, which was a fact about Edinburgh's geography rather than a plan.

November Edinburgh at ten o'clock: rain that was more mist than water, the orange of streetlamps on wet pavement, the particular cold that came off the sea and settled in the city's stone. He had been walking home through this weather for four years and had developed the Edinburgh-resident's particular relationship with it — not affectionate exactly, but accommodating, the way you accommodated something that was simply a condition of being where you were.

They walked without talking for a block, which was comfortable because they had been not-talking with each other for years. This was one of the things Sicheng valued most about Wei-Lin: she did not fill silence because silence made her uncomfortable. She filled it when she had something to put in it, and otherwise she walked beside you and let the night be what it was.

He had not had many friendships like this in Edinburgh. He had not had many friendships like this anywhere. The particular kind of trust that allowed for silence required a specific history, and most of the history he had accumulated in Edinburgh was professional rather than personal. He had been, by choice and by inclination, better at the former than the latter.

"A friend of mine just decided to stay in the UK," Wei-Lin said. "She was going to go back to Taipei after her postdoc. But she's staying."

"Because of the job?"

"Partly." Wei-Lin navigated around a wet patch on the pavement with the automatic efficiency of someone who had memorized Edinburgh's hazards. "Her partner is here. British. It's not — it's complicated, for her family. But she decided some things weren't about nationality."

He walked beside her and did not say anything.

"She said —" Wei-Lin paused in the way she paused when she was finding the right words for something. "She said she'd spent a long time thinking about what she owed to various things. Her family, her country, her career. And she realized she'd been putting the people she'd chosen to be with at the bottom of the list. Not because they mattered less. Because she thought the other things were clearer obligations."

The rain fell. Their footsteps made the sound footsteps made on wet stone.

"She's happy," Wei-Lin said. "Complicated, but happy."

He understood that Wei-Lin was not talking about her friend, or not only about her friend. He understood this in the way you understood things that were being said sideways, aimed at you without requiring you to acknowledge being aimed at. Wei-Lin had been his friend long enough to know that direct questions produced managed answers and that information delivered as story arrived somewhere closer to the actual thing.

He said: "I'm glad she's happy."

Wei-Lin glanced at him. Something in the glance said: *I know you understood what I was saying.* Something in his response said: *I know you know.* This was how they had always communicated about things that were real — with the particular efficiency of people who trusted each other enough not to require explicit confirmation of every step.

They turned onto his street.

"Six weeks," he said.

Wei-Lin looked at him.

"Until submission," he said. "The thesis."

"I know what you meant," she said. Not unkind. Just accurate.

They said goodnight. He went inside.

---

The flat was quiet in the way flats were quiet on Friday nights when you had come home from dinner and the week's work was nominally complete and there was nothing requiring your attention until Monday. He made tea he did not particularly want, sat at his desk, opened his laptop.

The shared document had a new entry. Timestamp: 8:34pm — while he had been in the restaurant, eating soup, monitoring his register, listening to Wei-Lin talk about a friend who had decided that some things were not about nationality.

*Tested the directionality model against the open system dataset. It holds. Better than I expected — the boundary conditions are more robust than the original framework predicted. Notes in section 4.*

He read it. He read it again.

*Better than I expected.* She did not typically use language about expectation in professional notes. She used language about evidence: *the data indicates, the results suggest, the framework predicts.* Expectation was a different register — it was about what she had carried into the test rather than what the test had produced. She had chosen to say *better than I expected* rather than *better than the model predicted*, which was a distinction he was aware he was parsing very carefully for a note about network boundary conditions.

He typed: *Good. I'll review section 4 tonight.*

He did not review section 4 tonight.

He sat with the shared document open and thought about Wei-Lin's friend who had decided some things were not about nationality, and about Wei-Lin looking at him across the restaurant table when he said *she's very good* without any register management, and about Daniel saying *Taiwanese people and mainlanders don't really mix here* in the same tone that Oliver had said *they've got a thing with mainlanders* and the fact that in both cases the observation had been accurate as a statistical matter and had felt, in the specific context of his own experience of the past eleven weeks, like a description of a general pattern to which he was apparently constituting an exception.

He thought about six weeks.

Six weeks until submission. Submission was not the end — there would be corrections, the viva, the administrative process — but submission was the moment the thesis stopped being a present obligation and became a completed thing, and after that the question of what came next would require actual decisions rather than deferral. He had been deferring certain decisions for four years. He had been deferring specific decisions for eleven weeks. He had been deferring the most specific decision for approximately four years and eleven weeks, which was, when he looked at it directly, quite a long time to defer something.

Wei-Lin had said: *she'd been putting the people she'd chosen to be with at the bottom of the list. Not because they mattered less. Because she thought the other things were clearer obligations.* He sat with this sentence for a while. He recognized it. He had been operating by a version of this logic for four years, and the version had seemed correct, and the correctness had been real, and he was now in the position of someone who had made a correct calculation and was arriving at the conclusion that the correct calculation had produced the wrong result.

He opened a new section at the bottom of the shared document, below the methodology notes and the parameter lists and the two lines that said *I've been thinking about it* and *So have I*. He typed:

*The methodology is stronger than either framework alone. That's not an accident.*

He looked at what he had typed.

It was not, strictly speaking, a professional note. It was not, strictly speaking, about the methodology. It was the kind of sentence that could be read as being about the methodology by someone who chose to read it that way, and could be read as being about something else by someone who knew how to read what he actually meant.

She would know how to read what he actually meant. He had been aware of this for some time.

He saved the document. He closed the laptop. He sat in the quiet Edinburgh flat with the rain on the window and thought about six weeks and about what it meant to have been calculating the wrong thing for four years and about the specific quality of *better than I expected* written by someone who used evidence language and had chosen, this once, to use expectation language instead.

Outside, Edinburgh kept raining. The shared document was saved. The sentence was there, waiting to be found.

He thought about Wei-Lin walking back to her own flat on the other side of the city, probably not thinking about him, probably already thinking about her simulation. He thought about *she'd been putting the people she'd chosen to be with at the bottom of the list* and about the specific architecture of how he had organized the past four years, and about the possibility — which had been growing from hypothesis to probability over the past eleven weeks — that the architecture was missing something structural.

Not wrong. Not wrong in the way the static directionality assumption had been wrong. But incomplete in the way that a framework built for closed systems was incomplete when applied to something that had always been, despite his best efforts at management, an open one.