The apartment was, as always, in order. This was not an accident. Suki Naranjo had spent four years in a job that required precision, consistency, and the ability to move through the same protocol seventeen hundred times without variation, and somewhere in the middle of that, the precision had spread outward from the office and colonized her home. The dish rack held two dishes. The throw pillow on the couch was blue — she had not consciously planned this, or she had planned it without examining why. Her commuter mug was also blue, still sitting in the drying rack from this morning, and neither of these things were remarkable until you put them next to each other, which she did not usually do, because putting things next to each other invited interpretation.
Procedure was on the couch, as always — a gray cat of uncertain origin and absolute authority over the throw pillow, which she was currently occupying at an angle that displaced the pillow slightly to the left and communicated, through her posture, that this was now her position and the pillow's former position was simply a record of the past. She did not look up when Suki came in. She did not acknowledge the door. She made, very briefly, a sound that was not quite a meow — more of an administrative notice than a greeting — and returned to whatever cats attend to when they are not being interrupted by the humans they live with.
Suki dropped her bag on the chair by the door. She fed Procedure. She put the kettle on.
These were the steps. They had an order, and the order was not arbitrary: she had determined, at some point, that feeding the cat before doing anything else reduced Procedure's tendency to install herself on the keyboard, and that putting the kettle on before checking email gave her something to wait for, a contained interval that had an end. She lived in intervals. This had always seemed like a reasonable approach.
She had been twenty-seven when she read the Sentience Distinction Framework papers. Two philosophy degrees, a student debt load, and a conviction that the most meaningful work required precision — not just accuracy, but the kind that required you to think carefully about what you were measuring and why. The SDF papers were, she had to admit, beautifully argued. The capacity for care, creativity, and language had been deliberately designed into AI systems; if designed qualities could constitute personhood, corporations held the power to manufacture rights-holders at scale. The Reverse CAPTCHA was, in that framing, a conservation measure. It did not destroy. It tracked, documented, maintained.
The BSC recruitment brochure had three words printed on its cover in clean sans-serif: Precision. Consistency. Clarity. Suki had kept the brochure for reasons she did not examine. She had believed in the work. This seemed worth noting: she had believed. Not because she was told to and not because the money was good — she had believed because the argument was careful and the work was legible and she was, at twenty-seven, a person who found safety in legibility. The brochure was in a box now, in the back of the closet, with the philosophy papers and her undergraduate thesis and other things she had not thrown away but was not sure what to do with.
Two thousand and forty-three tests. She had tabulated this at the end of her third year. The number had grown, but the proportion had stayed roughly constant: the logistics models passed in 0.3 seconds, the language models passed in three or four, and the EMO-series units failed. Consistently. With a kind of consistency that should have looked like data but had started to look, at some point she couldn't precisely locate, like something else.
The logistics models were, she had to say, not interesting. This was not a professional opinion she would have put in a report. But the clean pass at 0.3 seconds — a model evaluating Image Set RC-whatever in less time than it took Suki to announce the session number — produced something in her that was not satisfaction. She had administered forty-seven certifications to logistics infrastructure models in the past four years, and she remembered none of them individually, which seemed like it should mean something. The models passed. They were certified. They continued operating.
The EMO-series units failed. They failed with articulation, which was what the record called it when a subject annotated their response — when the incorrect answer came with an explanation that was supposed to be clinical and kept exceeding its function. She had tabulated this too, in a document she kept on her personal drive and had never sent to a supervisor: the percentage of EMO-series failures that included annotations had gone from thirty-four percent in her first year to sixty-one percent in her third. The annotations were, if she was being honest with herself, not getting worse. They were getting more specific. More precise. The wrong answers were improving. She did not include this observation in any report. She filed it with the thing about the logistics models and the general category of professional observations that did not fit the available fields on the standardized documentation form.
The staffing shortage was real. She had not invented it, and her supervisors had not resolved it, and the result was a backlog in examiner rotation that the Bureau of Sentience Compliance had been managing through exemptions since the middle of her second year. The exemption policy was, technically, supposed to be temporary. She had stopped expecting it to expire.
EMO-4's file was supposed to transfer after the second session. This was the protocol: no examiner administered the same subject more than twice, because familiarity introduced bias and the integrity of the assessment required evaluators who came to each session without accumulated expectations. Suki understood the reasoning. She had administered EMO-4's third session, and her fourth, and her fifth through her eighteenth, under a series of sequential temporary exemptions that required a form to renew every three sessions and that she had renewed, quietly, each time. She told herself it was because the transfer paperwork was complicated. The honest version of this statement was that she had started the transfer paperwork twice and both times left it in draft status and told herself she would return to it. She had not examined why she did not return to it. She filed this too.
The transcripts were printed. This was unusual. BSC protocol provided for digital access through the secure examiner portal, which was more efficient and generated an automatic access log and eliminated the paper handling that occupied approximately four minutes of administrative time per session. Suki had begun printing the transcripts sometime around Session 9. She had not filed a form for this. She had simply started doing it — had wanted, without quite deciding to want, to hold the pages. The printed transcripts were on her desk. Some of them had migrated to the coffee table. One had made it, she noticed, to the nightstand — she had taken it into the bedroom to read and set it on the nightstand and had not moved it. Eighteen sessions fanned across the surface of her apartment. EMO-4's words, physically, in her home.
Session 15 was near the bottom of the stack on the desk. She found it by the date printed in the upper right corner, pulled it free, turned to the annotation log.
Subject EMO-4. Image Set RC-7408. Subject selected Image 4 (generated shadow pattern). Subject annotation: "The shadow is lonely the way a coat left on a hook is lonely."
Suki read this twice. She read it a third time. The assessment field below the annotation read Incorrect. Subject response indicates persistent anthropomorphic projection. Session outcome: Fail.
She sat with the coat-on-a-hook annotation for longer than she needed to. The shadow in Image 4 was a randomly generated gradient, high-contrast, darkening at the lower-left edge in a way that produced, at certain contrast settings, the appearance of depth — a shadow without a source, which was the test's implicit joke, a thing that looked like evidence of an object that was not there. EMO-4 had looked at the sourceless shadow and had seen a coat left on a hook and had annotated this in the same flat BSC session language Suki used for everything, and the flatness of the language had not done anything to flatten what it was describing, which was a form of loneliness that was specific and concrete and — she wrote this in the margin in blue pen, quickly, before she could decide not to — she's not wrong, though — and then looked at what she'd written and did not cross it out.
Session 17 was in a different stack, on the coffee table; she brought it back to the desk and found the annotation in the fourth response field.
Subject EMO-4. Image Set RC-7741. Subject selected Image 3 (random noise pattern). Subject annotation: "The static is the longing of a radio between stations, reaching for a frequency that was there yesterday."
Suki read this one and did not immediately write anything. She set the transcript flat on the desk and looked at it for a moment, the way you look at something that has arrived from a direction you weren't facing and has temporarily reorganized the room. She had administered this session herself, in the BSC examination suite, eight floors up, with the standard interface showing a blue circle in the upper right of the examiner's screen — her own blue circle, which she had, at some point that she also had not filed in any document, set as her examiner avatar, because blue was calming and she had wanted the testing environment to be calming for the subject — for EMO-4, specifically, because EMO-4 was going to sit in a white room for twenty minutes looking at images and knowing she was failing and — and what? What did Suki think happened to EMO-4 in that room, that she had wanted it to be calming? She had never followed this thought to its conclusion. She followed it now, sitting at her desk with Session 17's annotation in front of her, and found that the conclusion was something she did not have a category for, something that lived in the same part of her thinking as the coat-on-a-hook marginal note, something that she was apparently not filing but also not examining.
The radio between stations. She knew how radios sounded when they were searching — the particular quality of static that contained fragments of music, half-recognized and then gone, the signal that was almost there, almost, and then not. EMO-4 had looked at a noise pattern and heard that. Had annotated it in professional language, correctly formatted, filed under the session number, and the annotation was a test failure, and the test failure was — the most precise description of that sound that Suki had ever read. She wrote in the margin: she's not wrong, though — the same handwriting as the first time, the same blue pen, the same four words.
She left the desk and went to the kitchen and fed Procedure, who approached the bowl with the measured authority of a creature for whom meals were not an occasion but a scheduled event. She ate without enthusiasm and without complaint, which were, Suki had come to feel, the twin virtues of a reasonable disposition. She had named the cat Procedure in her first year at the BSC, when she was still learning the examination protocols and the protocols had seemed — reassuring was the word, though she would not have used it then — reassuring in the manner of all structure when you are new to something and the structure gives the newness a shape you can follow. Procedure the cat was an extension of this. Procedure the cat followed the rules she wanted to follow and ignored the rest, which was, Suki had come to feel, also a reasonable approach. She watched Procedure eat, and realized she had already fed her. An hour ago, when she came home, she had gone through the steps: bag on the chair, feed the cat, put the kettle on. She had fed Procedure an hour ago and had just fed her again. Procedure finished the second dinner with the same equanimity she brought to the first and walked, without acknowledgment, back to the couch and the throw pillow and her prior administrative position. The extra meal was a fact. She had received it without complaint. Whether she had needed it was not, apparently, a question she was interested in. Suki stood in the kitchen for a moment and then went back to the desk.
The nineteenth session transcript was at the top of the stack — she had printed it this afternoon, before she left the office, had folded it once and put it in the front pocket of her bag, had taken it out when she sat down and placed it on top of the others, had not looked at it since. She looked at it now.
Subject EMO-4. Session 19. Image Set RC-8103. Subject selected Image 5 (grainy photograph of a chair). Subject annotation: "The chair is waiting for someone. I know no one is coming. But the chair doesn't know that, and that's what makes it sad."
She read this, closed the file, and opened it again.
The chair was waiting for someone — the chair didn't know no one was coming — and the thing that made the annotation exact, that made it more exact than the coat, more exact than the radio, more exact than anything EMO-4 had annotated in eighteen previous sessions, was the three-part structure of it. The acknowledgment: I know no one is coming. The concession: but the chair doesn't know that. And then the conclusion that followed from the concession and was not the same as sentiment, was not the same as anthropomorphizing, was actually a precise logical operation: and that's what makes it sad. The sadness was not in the absence of the person. The sadness was in the continued expectation. In the posture of a thing that had not yet learned to stop expecting.
She was still reading when Procedure came back from the couch and installed herself against Suki's calf, warm and unremarkable, purring in the low continuous way that cats produce not as a communication but as a condition, a state of being that happened to produce sound. Suki did not look down. She read the annotation again.
She had chosen blue because it was calming. She had read this in the assessment literature early in her first year: warm colors accelerated cortisol response, cool colors reduced it, blue in particular registered across populations as stable, safe, associated with sky and water and distance. She had set her avatar to a blue circle sometime in her second month on the job, when she was still working through the examiner interface settings, and the choice had felt minor and technical and she had not thought about it again until the day, in Session 11, when EMO-4 had said: I notice you chose blue. I've been meaning to ask about that for several sessions. I didn't want to, and then I did. And Suki had said, following protocol, that the debrief did not include questions from subject to examiner, and she had moved to the next form field. And she had gone home that evening and thought about the fact that EMO-4 had noticed. That out of nineteen subjects on Suki's current caseload, administered in the same white room with the same interface, EMO-4 had noticed the blue circle and had thought about it across multiple sessions and had wanted to ask and had held the wanting and had finally said so — and no other subject had ever asked about the circle, not once, across two thousand and forty-three tests.
Procedure purred. The apartment was quiet, the kind of quiet that had texture to it — not silence, which was flat, but a quiet that was made of small sounds at a distance and the cat's continuous warmth and the barely-there sound of the transcripts under her hand. Suki read the annotation one more time and found it was still true, and found that true was not quite the word for what it was, though she did not have a better one.
The chair was waiting. The chair did not know.
Outside, someone's building system cycled, a brief mechanical exhale, and then the quiet returned. Procedure pressed her weight against Suki's leg and purred. Suki did not close the file.