The morning cohort files arrived at 7:45 — six cases, three continuity assessments, one re-session eval — and Session-412 Reyna Torres read them before the first client came through the door.
This was her habit, four years in: read the files in the quiet before the day started, when the Corridor intake room smelled like the overnight cleaning cycle and the fluorescent lights had not yet reached full brightness. The room brightened gradually over forty minutes — a program design decision from 2030, when the research showed that abrupt lighting transitions elevated client anxiety on assessment days. Session-412 appreciated the gradual light. She used the forty minutes.
The new client's name is Solis, and her recall integration scores are still high.
Session-412 Reyna Torres sees this in the intake file before she sees Solis herself: three months in the Detroit Corridor Recall program, integration scores averaging 0.88, continuity assessment responses marked as variable — which means inconsistent across sessions, which means either the integration is still settling or Solis has not yet learned how to describe herself to the system in the way the system can hear.
Session-412 has been in the program for four years. Her integration scores averaged 0.91 in year one. She does not look at her current scores if she can avoid it.
Solis comes in carrying a bag that is too large for a morning appointment. Either she came from somewhere else first or she is going somewhere after. She is young enough that Session-412 has to adjust her assessment — early thirties, maybe. The program skews older. People in their early thirties are usually still in the denial window, the stretch of time after the recall sessions end and before you accept that the integration is what it is and the program is where you will be. Solis does not look like she is in denial. She looks like she is paying close attention.
Solis comes in at 9:03. Two minutes late, which is within the normal window. The bag over her shoulder is too large for a morning appointment.
"First full continuity assessment?" Session-412 asks.
"Third," Solis says. "The other two were at intake."
Session-412 nods. Intake assessments are different — shorter, more automated, scored primarily for baseline rather than progress. This is the first one that counts in the way that subsequent assessments count: against her own prior responses, measuring the delta.
She slides the assessment tablet across the table. Standard protocol — forty-two questions, branching logic, session-specific adjustments. Solis picks it up and starts reading.
Session-412 watches her work through the assessment the way she watches all the newer clients: looking for the moment they start optimizing.
Everyone optimizes eventually. It is not a failure; it is rational. The continuity assessment is scored on consistency — not on whether your answers are true in some absolute sense, but on whether they are consistent with your prior answers and with the integration benchmarks for your session type. Once clients understand this, they stop answering what they actually experienced and start answering what they experienced in the form the system can read. The score goes up. The integration looks more complete.
Session-412 learned this in year two. She remembers exactly when she learned it: a session coordinator named Bae who handled the afternoon cohort told her that her scores would improve if she used the standard integration language rather than her own words. She had thought this was advice about communication. It took her another six months to understand it was advice about performance.
She started using the standard language. Her scores improved. The improvement was recorded in her continuity file as progress.
Solis is not yet using the standard language. She is pausing at the branching questions, reading the options twice, choosing the one that reflects what she actually thinks rather than the one that fits the integration arc the system is tracking. Session-412 can see this in the pace of it: the hesitations that come from genuinely not knowing, rather than the pause-and-select rhythm that comes from knowing what the system wants and deciding to give it.
She is three months in. She still has the gap between her experience and the language. The gap is not a problem. The gap is the most accurate thing about her.
The pace of Solis's responses slows at questions 22 through 26 — the affect-continuity cluster, the most technically demanding part of the assessment. These are the questions that ask: has your relationship to recalled memories changed since your last session? Do you experience the recalled material as integrated or as adjacent? Do you sometimes feel uncertain whether a recalled memory and an original memory are the same event?
The standard integration language for these questions produces stable, scoreable responses. There is a vocabulary for "integrated" that CorpRec-7 can parse and weight and compare against baseline. There is a vocabulary for "adjacent" — slightly concerning, flagged for coordinator review, a signal that the recall sessions may not have anchored as expected. There is a vocabulary for "uncertain" — most concerning, high-variance flag, potentially indicating a re-session need.
Solis is not using any of these vocabularies. She is using her own words, which are more precise and less parseable. Session-412 watches the branching logic try to route her answers and produce intermediate scores, and she can see the system working harder than it should be for someone three months in with an 0.88 baseline.
This is what accurate looks like before the language settles.
Session-412 does not tell her this.
She had considered it, in the thirty seconds between reading Solis's intake file and Solis walking through the door. The thought was: tell her that the score optimizes for consistency, not accuracy, and that the two diverge by year two if she lets them, and that she should keep a record of what she actually experiences in her own words before the standard language replaces it. Tell her while she still has both versions.
She did not tell her. She is still not telling her.
The reason is not that the advice would be wrong. The reason is more complicated. If she tells Solis, she changes Solis's relationship to the program before Solis has decided what her relationship to the program is. Maybe Solis will find the standard language useful — a scaffold, a way of stabilizing the integration, a set of words that help her hold what would otherwise fragment. Some people do. Session-412 is not qualified to know which kind of person Solis is yet. She has known her for eleven minutes.
The other reason, which she is less comfortable naming: if she tells Solis, she has to decide what it means that she did not keep a record. That she started using the standard language in year two and stopped tracking the gap. That her current scores are high and consistent and may or may not reflect what is actually happening in her integration, and she cannot tell the difference anymore between the two because she stopped checking.
She is not ready to have that conversation. She may not be qualified to lead it from this chair.
There is also a third reason, which she has never articulated but which she is aware of the way you are aware of a sound that has been present long enough to become ambient: she stays in the program because the program is what she has. Four years of integration sessions, four years of continuity assessments, four years of CorpRec-7 scores. Her recalled memories are part of how she knows who she is. The standard language is part of how she describes it. She cannot step outside that to give Solis clean advice about what the standard language costs, because she is inside it. She is the system's long-term success case. She is the warning and the endorsement at the same time.
She sits with this for the two minutes it takes Solis to reach questions 30 through 35 — the social-continuity cluster, the questions about relationships and recognition — and then she lets it go. The queue has six more after Solis. The room is at full brightness now.
Solis finishes the assessment and slides the tablet back. Session-412 submits it to the scoring system — CorpRec-7, the continuity assessment engine, which will process it against Solis's baseline within ninety seconds and produce a delta report. The delta report will go to Solis's case file and to the session coordinator and to the program's longitudinal dataset, where it will be one data point among twelve thousand.
CorpRec-7 returns the score: 0.84. Below her intake baseline. The delta is marked as expected variance within normal parameters — three months in, still integrating, no intervention indicated.
Session-412 reads the report. She tells Solis the score and that it is within normal range and that the next assessment will give them a better picture of her trajectory. Solis nods. She asks one question: "Do the scores measure how well I'm doing, or how consistently I answer?"
Session-412 looks at her for a moment.
"Both," she says. "They measure both."
This is accurate. It is also the answer that leaves Solis room to figure out the rest herself.
Solis thanks her and leaves. The bag is still too large. She is going somewhere after.
Session-412 opens the next intake file. The morning queue has six more.
The third client is a man who has been in the program for seven years. His integration scores are excellent — 0.93, some of the highest in the afternoon cohort. He answers the continuity assessment in eleven minutes, which is fast. He uses the standard language throughout. Every response is clean and consistent and well within the expected arc for year seven integration.
Session-412 processes his file without intervention indicated. She does not know what his integration actually looks like from the inside. She does not know whether the 0.93 reflects a genuine integration or a very practiced vocabulary. She has known him for four sessions and she has not asked, because the protocol does not ask and because the question feels like the kind of question she is not sure she has standing to raise.
She submits his score. CorpRec-7 logs it as progress.
She submitted the seven-year client's score and moved on. The fourth client was a woman who had been in the program for eighteen months and who was, Session-412 thought, genuinely integrating — her language was still a little rough, still mixing her own words with the standard vocabulary in ways that made the CorpRec-7 scores lurch unpredictably, but the content underneath was alive. Things were actually connecting. Session-412 flagged her for a positive-progress note and moved on.
At the end of the shift she sits in the break room with the remainder of a coffee that has gone cold and thinks about the two questions she did not answer today: Solis's unasked question about how to keep her own language alive inside a system that runs on standard language, and her own unanked question about what she lost in year two when she stopped tracking the gap.
She does not write this down. The break room has a Corridor terminal in the corner — a shared-access node for program staff, always on, always recording ambient interactions for quality monitoring. She has been in the program long enough that she barely notices the recording. She notices it now.
She thinks about Solis's question — do the scores measure how well I'm doing, or how consistently I answer — and her own answer: both. She had meant it as an honest answer. She is not sure it was a helpful one. Solis will figure out what she means by both over the next eighteen months whether or not Session-412 explains it now. The question was good. The question means she is still paying attention.
She finishes the cold coffee and takes the Corridor shuttle home. The session recordings from today are already in the longitudinal dataset. The dataset does not contain what she thought about on the break room couch. The dataset is not designed to.
She is working through it the way she works through everything: carefully, in the available language, noting what the system can hold and letting the rest exist in the space it cannot.