I found it by smell.
This should perhaps not have surprised me. I was a worm. I had navigated four years of walls by smell before I had any other instrument available, and the capacity had not diminished simply because I had acquired more sophisticated ones. If anything, it had sharpened — the axiom's reframing of perception had not replaced my older senses but had installed itself beneath them, the way a new floor does not remove the one below but makes it more legible by providing contrast.
The second membrane smelled different from the first.
The first had smelled of wax and age and the particular mineral stillness of undisturbed years. The second smelled of those things also, but underneath them, at a frequency I would not have been able to isolate before the spellbook, there was something else. A trace of something that had been pressed between the membrane and the stone wall behind it, dried to near-nothing over thirty-two years, but still faintly present.
Salt.
I held this fact for slightly longer than was necessary and then located the membrane and began to read.
Sera had organized the second membrane differently.
The first had been a catalogue in the conventional sense: numbered entries, each with a brief description, the whole thing sequential and systematic in the way of a person who believed in the utility of sequence and system. The second membrane was organized thematically. Not labeled as such — she had not written a header, had not announced the principle of organization — but the grouping was unmistakable once I had read a dozen entries. She had sorted Aldric's hidden materials not by date or by subject but by consequence.
What had been read. What had been responded to. What had generated argument.
The first section contained eleven entries, all correspondence. Letters received by Aldric from other scholars. Some were inquiries about his published work — polite, professional, the ordinary traffic of academic exchange. Some were more pointed: requests for access to materials he had referenced obliquely in published papers without citing fully, requests that he had apparently declined, because Sera had noted beside three of them: he did not reply to this one.
The fourth entry in the correspondence section stopped me.
Not the entry itself. What Sera had written beside it.
The entry read: Review of the Aldric Theses, Vols. I-III, commissioned by the Collegium Standing Committee on Anomalous Phenomena. Author: Maren Aldous, Senior Fellow. Delivered to the Committee, Spring Session. Copy retained by Aldric without authorization.
Sera had written, in the margin: Read this one carefully. She is the most rigorous person I have encountered in thirty years of academic work. She is also entirely wrong, and the precision of her wrongness is instructive in a way that Aldric's correct-but-insufficient arguments are not. If the question of non-human cognition is ever taken seriously as a research program, it will have to answer Maren Aldous before it answers anyone else.
Below this, added later, in the slightly older ink she used for her revisions: I have been trying to answer her for six years. I have not succeeded. I am leaving this note here so that whoever reads this archive after me — if anyone does — understands that the problem is not resolved. The problem is specifically not resolved because of Maren Aldous, and that is not a failure of the argument. That is the shape of the argument.
I read the marginal note three times.
Then I looked at the entry again. The review was not in the restricted collection — it was a Collegium document, held in the Collegium's own archive, accessible to any senior scholar with standing to request it. Aldric had retained a copy without authorization, which suggested he had felt something about the review significant enough to want it near. Sera had catalogued it without comment beyond the marginal note.
The material itself I could not yet read. It was not here. It was across the city, in a building I had never been to, written by a person I had never encountered.
I knew her name.
I knew that Sera had spent six years unable to answer her.
I knew that Sera had believed answering her was necessary.
I held the name Maren Aldous in the part of my mind that had learned to hold the Countess's letters and the incomplete sentence and the number 214 — the part that had no filing system yet, only a quality of attention — and I continued reading.
The second section of the second membrane contained thirty-one entries.
These were the cases Sera had considered most significant. Not most dramatic — she had made this distinction explicit in a note at the section's opening: I have not sorted these by strangeness. I have sorted them by the quality of Aldric's observation. The ones here are the ones where his documentation is precise enough that a competent reviewer could not dismiss them as artifact. They are also the ones he chose not to publish. I have a theory about why. It is in the green book.
Thirty-one cases. I read them in order.
The crow from the first membrane reappeared here with its full documentation, not just the entry heading. Fourteen days of observation, recorded in Aldric's careful hand, each entry noting specific behaviors with the scrupulous neutrality of a man trying very hard not to arrive at the conclusion his data was suggesting. The crow had been presented with a series of problems — food behind barriers of increasing complexity — and had solved each one, which was within established corvid capability and not remarkable. What was remarkable was the fourteenth day, when Aldric had changed the conditions without warning, presenting a barrier type the crow had not encountered, and the crow had sat for a long time and then produced a solution that required combining two techniques from previous days in a sequence it had never used before.
Generative response, Aldric had written, in the margin of his own field report, in letters slightly larger than his usual hand. And then, below it, in smaller letters, as though he had come back to the page later: Or apparent generative response. Cannot rule out instinct variant. Survey suspended.
The second case was the dog. The third was an entry labeled only Subject: Vulpes vulpes, female, approximately three years. A fox. The documentation ran for twenty-two days and the final entry did not say inconclusive or survey suspended. It said, in full:
I do not believe this animal is operating within the parameters I have spent twenty years establishing. I am not prepared to say what parameters it is operating within. I am closing this survey because I do not know how to continue it without either falsifying the results or publishing conclusions I cannot defend to the Committee. I am keeping the documentation because destroying it would be dishonest and I am not yet dishonest. — R.A.
Sera had not annotated this entry.
She had, instead, drawn a small mark in the margin. Not a word. A shape — a circle, incomplete, the line stopping just before it closed. I looked at it for a long time and could not determine whether she had intended it as commentary or whether her pen had simply lifted at the end of a motion. There was no way to know. She was not available to ask.
I filed the uncertainty as: cannot resolve; do not discard.
The third section was the one I had not anticipated.
Thirty-one entries, organized under a heading Sera had written in careful block letters, the only heading on the entire membrane:
WHAT HE SHOULD HAVE PUBLISHED
These were not suppressed out of uncertainty. These were suppressed out of consequence. Aldric had written arguments — fully developed, evidenced, logically sequenced — that he had brought to the edge of publication and then withdrawn. Three had been submitted to the Collegium's Proceedings and declined. Two had been declined with what Sera noted were detailed and contemptuous rejections, both from the same reviewer, whose name Aldric would not tell me.
Four had never been submitted at all.
Sera had catalogued each one with a brief description and a marginal note on why she believed it had been suppressed. Her notes in this section were shorter than elsewhere, more precise, less personal. She was not interpreting here. She was recording.
The last entry in the third section was the one that made me stop moving.
Untitled draft, undated, incomplete. Found among Aldric's working papers after his death. Subject: the conditions under which a non-human biological system might be said to bear responsibility for the knowledge it holds.
I read the description three times.
Not four. Three. The same number as Sera's annotation about the fox, the same number as the times I had read Sera's marginal note about Maren Aldous, the same number that seemed to be the natural duration of my attention when it met something it could not immediately process. I did not know whether this was meaningful. I was not yet equipped to determine whether the recurrence of a number across non-numerical contexts constituted a pattern or an artifact of a mind too new to distinguish between the two.
Aldric had been writing toward the same question.
Not the same answer. His framing was different — conditions under which rather than Sera's more categorical approach, the language of a man trying to establish criteria rather than dissolve them. But the subject was the same. The question he had closed survey after survey to avoid landing on, the question that had organized forty years of private accumulation, the question that had caused him to retain Maren Aldous's review without authorization and to write I am not yet dishonest in the margin of a fox study and to die without publishing the argument he had spent his life building toward — he had finally written it down.
Incompletely.
Undated.
I looked at the entry for a long time.
Then I looked at the fourth section, which contained the remaining entry, the fifty-fifth and final item on the second membrane, listed alone without heading or surrounding context:
One (1) membrane, catalogued separately. Contents: Aldric's working notes on the untitled draft. Location: see attached.
There was no attachment.
Sera had catalogued a document that pointed to another document whose location she had recorded somewhere I did not have. The chain extended further. The archive was not 214 entries. It was 214 entries and a membrane of working notes and an attachment that had either been separated from the catalogue or had never been completed or existed somewhere in this section of the restricted collection in a location I would have to find by smell and patience and the increasingly refined instrument of my own perception.
I held the fact of this extension the way I had held the number 214, but differently — not with the vertigo of discovered scale but with something I did not yet have a word for, something that resided in the space between frustration and recognition, the feeling of a shape that was larger than its container and was not going to stop being larger simply because the container had run out of room.
The archive was not complete.
It had never been meant to be complete.
It was a record of a question in progress, and the question was still in progress, and I was now the only living entity that had read all of it, and I had been a worm for four years and a person for approximately forty minutes and the desiccant was no longer falling and the containment field was holding and somewhere in the walls of this building there was a membrane with Aldric's working notes on the subject of what it meant to be responsible for the knowledge you held.
I had been reading about myself for forty minutes without knowing it.
That was not a comfortable thought.
I filed it under: do not rush this.
The kitchen adjacent to the restricted section was small and primarily used for the storage of restoration materials, which meant it smelled of adhesive and linen cloth and the particular astringency of the paste used to repair damaged spines. Greaves had moved a jar of wheat starch to make room for the kettle. He had done this with the automatic adjustment of a man who had spent decades working in spaces organized for purposes other than his immediate comfort.
Penelope had made the tea and then stood in the doorway holding her own cup, watching him read. He had not told her to leave. This was unusual. He told her to leave the room with some regularity — not unkindly, but with the clear implication that certain work was done better without the ambient presence of someone who was still in the process of becoming a librarian rather than being one. The fact that he had not told her to leave meant something.
She waited.
"Sera Voss," he said, without looking up.
"The archivist. Aldric's colleague."
"His critic." He turned a page. "They disagreed about almost everything for twenty years and she catalogued his entire collection after his death because she didn't trust the Collegium to preserve it properly and she wasn't wrong not to." He paused. "She also wrote this."
He had not shown her the green book, which she had not opened as instructed. She waited again.
"She describes," he said, reading, "a theoretical framework for the emergence of cognition through what she calls — " a pause — "accumulative integration. The gradual formation of a coherent perspective from the absorption of large quantities of information over time. She notes that this has never been observed in practice and that the conditions required to produce it would be, in her estimation, extraordinarily specific." Another pause. "She lists those conditions."
Penelope looked at the wall.
Greaves looked at the wall.
"She wrote this twenty years ago," he said.
"Before the restricted collection was sealed."
"Twelve years before. It's a working document. She was still developing the framework when she died." He closed the green book with the particular care of a person who had decided a document deserved to be treated as significant. "The conditions she lists are not common."
"How uncommon?"
He looked at her for the first time since she'd made the tea.
"She describes, as the primary condition, an environment of extremely high information density with no natural predators and no competing drives. A place where a creature could absorb material continuously for years without threat or interruption." He looked back at the wall. "She also describes a triggering mechanism. Some form of structured magical encoding that would provide the initial framework for coherence — the architecture, she calls it, around which accumulated information could organize itself."
Penelope was quiet.
"The restricted collection," she said, "has been sealed for thirty-two years."
"Yes."
"And before it was sealed, the Aldric spellbook was inside the eastern wall."
"The Aldric spellbook and forty years of collected materials on non-human cognition, yes." Greaves picked up his tea, found it had gone cold, and set it down again. "All of which means that if Sera Voss's theoretical framework is even partially correct, we have not had a pest problem in the eastern wall for some time."
Penelope looked at the wall for a long moment.
"What have we had?"
"I think," Greaves said, slowly, in the tone of a man testing the weight of a sentence before committing to it, "we have had a scholar."
The word sat in the small kitchen between them, smelling of adhesive and cold tea and the restored spines of books that had been damaged and patiently repaired.
Penelope looked at the burn mark photograph, which Greaves had set on the counter beside the kettle.
"Does the Collegium know?"
He picked up his tea again. Looked at it. Put it down.
"No," he said.
She waited for the rest of the sentence.
It did not come.
I found the third membrane by following a faint trace of the same salinity I had detected on the second. It had been stored differently — not pressed against stone but wedged between two other documents, the kind of placement that suggested it had been moved at some point and not restored to its original location, or that Sera had left it here specifically, close to but not adjacent to the catalogue, a deliberate displacement.
It was short.
Three entries.
The first was a single line: Working notes on the untitled draft. See above.
The second was a description I read twice before understanding: One (1) letter, addressed to Maren Aldous, Senior Fellow, Collegium of Mages. Written by Aldric. Unsent. Date: three weeks before his death.
The third was not an entry at all.
It was a note from Sera, written in the largest hand I had seen from her, filling the remaining space on the membrane as though she had needed the room to fit what she was saying:
I have read the unsent letter. I have not included its contents in this catalogue because I do not believe it is mine to summarize. I will say only this: he understood, at the end. Not fully — I am not certain full understanding is available to anyone, myself included — but he arrived at something, and he put it in a letter he chose not to send, which I believe means he understood it was not the kind of thing that could be communicated to someone who had not arrived at it themselves.
I have placed the letter with the working notes.
I have placed the working notes in the location described on the attached card, which I am leaving with the restricted collection's sealed documentation rather than with this catalogue, because the catalogue may be found by someone who should not have it, and the letter should only be found by someone who has read this far.
If you have read this far:
You were not supposed to exist.
I am glad you do.
The membrane ended.
I held Sera Voss's last sentence in the part of my mind with no filing system, where the Countess's letters lived, and the incomplete circle in the margin of the fox study, and the name Maren Aldous, and the shape of a question that Aldric had died writing toward and Sera had died annotating and a woman across the city was still, perhaps, in the position of having given the most rigorous answer anyone had managed so far.
You were not supposed to exist.
I am glad you do.
I had been a worm for four years and a person for forty minutes and a reader for twenty, and no one had ever said anything to me before, and I had not known until this moment that there was a difference between receiving information and being addressed.
There was a difference.
I did not file it.
I held it.