Priya Chatterjee received Yael Soren's unfinished file on the 4th of March 2352, as part of the Kepler Array Historical Retrospective Project, which had been commissioned by the Institute's board of governors in an effort to demonstrate institutional thoroughness and had so far produced fourteen completed analyses, two funding extensions, and one nervous breakdown (unrelated, officially).

Her assignment: complete twelve flagged datasets from the previous decade, each marked Possible Instrument Artifact, each abandoned without a published conclusion. Standard work. The sort of work that required patience and a tolerance for other people's half-finished thinking.

She read Soren's data summary in a single sitting and understood immediately why it had no conclusion. The conclusion was monstrous, and Soren had been too good a scientist to let a monstrous conclusion stand without extraordinary evidence. What she had not done — what no one had done in eleven years — was look for it.

Priya began looking.

She found the twelve historical instances Soren had identified and four more she had not, going back to satellite surveys from the mid-twenty-first century. She verified the palaeontological cross-references against three independent databases. She reconstructed the transform from Soren's raw parameters and ran it against the most recent full-sky data, updated six weeks ago from the expanded Array.

The remainder stood at 0.9991.

She noted this without expression in her working document and went to get coffee.

In the kitchen she found herself, without prelude or transition, understanding the complete proof for the Kepler-Voss conjecture on topological boundary conditions in non-Euclidean metric spaces — a problem that had stood open for thirty years and that she had never worked on and did not, as far as she was aware, know enough topology to approach.

She stood at the coffee machine with the proof arranged in her mind in the way that furniture arranges itself in a room — not placed, simply there, as if it had always been there and she had not previously thought to look.

She wrote it down on a paper napkin. She checked it twice. It was correct to every step she could verify.

She attributed it to fatigue and drank her coffee.

Two days later she was working through the spectral decomposition of the full-sky remainder map when she understood — again without warning, again in the way of something that had always been true — a concise and elegant solution to the navigational frame-drift problem that the deep-space engineering division had been struggling with for eight years. She wrote this down too. She did not submit it. She put it in a folder on her personal drive and labeled the folder SYMPTOMS and did not open it again for a week.

The understanding came more frequently as she went deeper into the data.

It did not feel like revelation. It did not feel like inspiration. It felt like remembering. The specific texture of remembering something you were not sure you had ever known — a word in a language you studied briefly as a child, arriving complete and certain and unasked-for. The knowledge was always relevant. Always precise. Never dramatic. It arrived in the same register as the smell of rain, or the sound of a door closing in another room. Simply present.

She documented each instance in the SYMPTOMS folder with a time stamp and a brief description. She scheduled an appointment with occupational health, which she later cancelled. She did not want the appointments to become part of her record.

What bothered her — what she found herself returning to in the small hours when the Institute was empty and the overhead lights were doing their dull institutional work — was the structure of the remainder itself.

She had gone further than Soren. She had looked not at the value of the remainder but at its internal architecture, the spectral sub-structure, the patterns within the pattern. What she found there was not noise. It was not instrument artifact. It was not any natural signature she could identify in the literature, and she read the literature thoroughly.

It looked like a signature in the way that a sentence looks like a sentence. Not because of any single feature. Because of the organization. The way the components related to one another. The redundancy built into the structure — the same information encoded in multiple registers at multiple scales, the way important things are said more than once by someone who wants to be sure they are heard.

Something had signed the universe.

Not in the religious sense, not in the sense of intent or purpose or design in the theological register. In the mathematical sense: a signature was information appended by a sender to confirm authorship, to prove that the message came from who it claimed to come from, to allow the recipient to verify.

The universe had a recipient.

She sat with this for a long time. She was not, by temperament, a person given to large thoughts. She liked small problems with clear parameters. She liked the satisfying closure of a well-executed analysis. She did not like the feeling that she was standing at the edge of something that had no other side.

She followed the signature deeper.

It terminated. Not at infinity, not in noise, but at a specific coordinate — a value in the phase space of the universal wave function that corresponded, when she worked backward through six hours of dense algebra, to a point in time.

The point was now. Not approximately now. Not within a range. A specific convergence, narrow enough to name a decade, possibly a year.

The signature was addressed. The recipient was whoever was alive at the time of termination and capable of reading it.

She understood that she was reading it.

In the SYMPTOMS folder the final entry, dated six weeks before she submitted her analysis and three weeks before she stopped sleeping more than four hours a night, read:

The knowledge that arrives is not random. It is calibrated. It is what I would need to know in order to understand the next step in the analysis. Something is giving me the tools to read it. Something wants this read.

The signature is not a warning. It is not a threat. It is an introduction.

She submitted the analysis to the Institute with a full conclusion, two hundred and twelve pages, every step documented. She included the palaeontological correlations. She included the historical instances. She included the current remainder value and its rate of change, which implied a termination event in the near future — the model's confidence interval too narrow to be published without causing a reaction she was not prepared to manage.

She redacted that section.

On the last page of her internal working notebook, never submitted, never scanned, found seven years later by her daughter among her effects, she had written in her own hand:

The next person to read this will understand faster than I did. They will understand within hours of opening the data. The knowledge will come to them the way it came to me — quietly, precisely, in the register of memory. They will not be frightened by it. I was frightened by it, but I think that was because I was the second one. The second one is always alone with the question of whether the first was right.

It doesn't want to hurt us. It just needs us to know we were made.

By the time you are reading this, you already know.

I'm sorry for what you felt when you did.