I.
The brief is ninety-four pages. Reyna has read it before — twice, during the deposition prep, and once more when the Holloway verdict came back. She read it the way you read something you are looking for a mistake in: fast, skipping toward the sections you already know are wrong, trusting your memory of where the problems are.
She has been wrong to trust that.
She found the sentence at eleven forty-three on a Sunday, which she will remember later as the particular indignity of the timing. Not during a working session. Not while building the Pre-Session Window proposal. While eating instant noodles in the certification district overflow housing she has lived in since the Tended sector rezoning, a brief open across her knees, looking for something to cite in support of paragraph seven.
Page 34. Torres v. RECOL Certification Board, 2028. Her name in italics. Standard citation formatting. The sentence below it reads: Pre-session behavioral observations, while potentially indicative, do not constitute certified extraction context under RECOL Procedural Code 12-A.
Reyna sets down her noodles.
She reads the sentence again. She knows this sentence. She has cited it herself in three separate consultation reports. The sentence is the current standard for what a Variance Log can and cannot establish in a RECOL-adjacent proceeding. The sentence is her testimony. She was there.
The word potentially is in the sentence.
She opens her attestation archive. The 2028 deposition runs to 220 pages; the relevant exchange is on page 84. She was asked whether pre-session behavioral observations could establish extraction context. She said: They can be indicative, in my experience. Whether a specific set of observations rises to the level of certified context is a separate determination. The transcriptionist coded the exchange. The RECOL Procedural Review Board distilled the exchange into a one-line standard. Somewhere in the distillation, can be indicative became potentially indicative.
It is a small word. The Board was not wrong to include it — it captures the epistemic hedge she intended. She was careful. She had spent twenty minutes in the deposition room choosing the hedge, because she did not want to overclaim, because overclaiming was how Variance specialists lost their certification, because she had seen it happen to Mara Singh three years before the testimony and she had paid attention.
She was a good witness. She hedged where she should hedge. And then she forgot she had done it, because she moved on to the next case and the next deposition and the seventeen other observations she had catalogued in seven years that she had not yet translated into formal language.
Page 34 of the Holloway brief uses the one-line standard to exclude Reyna's Variance Log from admissible evidence. The defense argument is eight sentences. The core of it: The Torres standard itself acknowledges that pre-session behavioral observations are 'potentially' indicative — not reliably so. Dr. Torres's Variance Log, regardless of its thoroughness, falls within the epistemically uncertain category her own testimony describes.
The argument is correct. That is the worst part.
II.
She opens the RECOL interface.
It responds in 0.3 seconds, which is the current standard under v3.8. Earlier versions responded faster — v2.9 could pull a session index in under 100 milliseconds — but the oversight revision after the Dakar incident added mandatory latency, a deliberate pause built into every data call to create what the Board called a moment of human orientation. Reyna has always thought this was theater. The pause does not reorient her. It makes her aware that the pause exists.
RECOL v3.8.2. Certification status: current. Last sync: 06:47. The interface shows Holloway's session record the way it always has: a green coherence bar at 97.3%, session duration forty-four minutes, the extraction quality notation that took up a quarter-page in the original report. Below the quality bar, the behavioral annotation section. Reyna's Variance Log, three pages, now marked in the archive as excluded — procedural 12-A.
She looks at the annotation she wrote in the waiting room before the Holloway session. Session candidate presents elevated baseline cortisol indicators per ambient biometric. Hands cease movement entirely at question three. Self-correction frequency higher than population average for this age cohort and extraction type. She wrote this in the twenty-minute pre-session window she had developed over years of informal practice — the window the Pre-Session Window proposal is trying to formalize and certify.
The paradox was not invisible to her before tonight. She had known, in an abstract way, that her own testimony created the standard that excluded her own methodology. She had discussed it with Dr. Nakamura at the February calibration conference. The Torres catch, Nakamura called it, which Reyna did not find funny. She had framed it as an obstacle to be legislated around, not a sentence she had written herself with twenty minutes of deliberate word choice in a deposition room in 2028.
She opens the proposal draft. Paragraph fourteen: Current RECOL procedural standards, including the Torres 12-A standard, treat pre-session behavioral observation as supplementary rather than constitutive evidence. This proposal argues for reclassification of a specific subset of behavioral data — sustained and replicable observation within the certified pre-session window — as primary extraction context. She wrote Torres as a reference without a note to herself that she is Torres. She has been doing that for two years.
III.
The RECOL system sends her an unsolicited prompt at 12:04 AM. This is new in v3.8 — the contextual consultation feature the Board announced as a quality improvement and the Variance Association called a surveillance mechanism. The prompt reads: You have been reviewing Holloway session archive materials for 22 minutes. Would you like to initiate a documentation session?
She says no. The system accepts this without argument, which she appreciates. Earlier versions might have asked twice.
What she is thinking about is not the proposal. She has been thinking about it in the wrong register. She has been thinking: the standard needs to change, here is the argument for changing it, here is the evidence. She has not been thinking: I was the person who chose the word potentially, and I chose it carefully, and it was the right word for what I believed in 2028, and what I believed in 2028 was accurate, and the Holloway defense is also right, and all of this is true at the same time.
She opens a new annotation file in her personal archive. Not the RECOL system — her own encrypted observation log, the one she has kept since year three of her practice when she started suspecting that the things she noticed before extraction began mattered more than the Board's methodology allowed for. She types:
Holloway. The annotations I wrote in the pre-session window were correct. The session coherence was 97.3%. The Variance Log I filed noted elevated baseline stress indicators, elevated self-correction rate, cessation of incidental movement at question three. Those observations predicted that the Holloway extraction would show high surface compliance and possible suppressed affect in the lower register — which it did. I was right. I documented being right. The documentation is not admissible because the standard for admissibility was written from my own testimony, in which I hedged, as I should have hedged, because I was being honest about the limits of what observation can certify.
The word potentially is not wrong. The problem is that the word potentially is being used to exclude documentation that was, in this case, correct. Correctness and certifiability are different things. I have spent nine years trying to turn the first into the second.
She reads this back. She thinks about Mara Singh's certification hearings, which she attended as a junior specialist and which ended Singh's practice not because Singh was wrong but because Singh claimed certainty she could not demonstrate. She had paid attention. She had learned. She had been so careful to hedge that she had, nine years later, contributed to the exclusion of her own work by a defense attorney who read her deposition more carefully than she had.
IV.
At 1:17 AM, the building's environmental system dims the overhead lights. RECOL district overflow housing runs on a tiered energy schedule under the Tended sector agreement — a concession extracted from the rezoning negotiation that Reyna had signed onto because it reduced her monthly certification infrastructure fees. She does not always notice the dimming. Tonight she notices.
She looks at the Holloway Variance Log one more time. The hands stopping at question three. The self-correction frequency. The elevated baseline. She documented these things in twenty minutes before a session she was then required to leave. The certified portion of her observation did not include what she saw — it included the instruments she used to see it. The RECOL v3.8 ambient biosensor data. The session intake form. The pre-session questionnaire with its standardized responses.
What the instruments did not capture: she had watched Holloway pause before answering question three for 1.2 seconds. This was not a long pause. She had written it down because she noticed it, and she noticed it because she had developed, over nine years, a specific attentiveness to pauses that began before the question finished. That attentiveness was not certifiable. It was hers.
The proposal needs a different argument. Not reclassify the observations — the Holloway defense would use the Torres catch against that approach too. Something prior: certify the window in which observation occurs. Not the individual annotations but the condition that produces reliable annotation. If the twenty-minute pre-session window is formally recognized as a certified observational context — if the practitioner's presence in that window is itself a certification act — then the observations produced in it carry different weight.
She has been making this argument, but she has been making it as an exception to the Torres standard. It should be a restatement of it.
She opens the proposal. Paragraph fourteen gets a line through it. She starts over.
The Torres 12-A standard correctly establishes that behavioral observation cannot replace extraction context. This proposal does not contest that standard. It argues instead that certified observational presence — a practitioner credentialed to RECOL-adjacent behavioral attestation standards, operating within a formally defined pre-session window — constitutes a form of context generation rather than context documentation. The practitioner does not record what exists. The practitioner's presence is itself an evidential act.
She reads this back. It is better. It does not undo the word potentially — nothing will undo it; the word is correct — but it steps around the problem by changing the question. Not: are these observations reliable? But: is the person making the observations certified to generate context?
At 1:47 AM, RECOL sends another contextual consultation prompt. You have been active in the archive for 64 minutes. Active documentation sessions are available. She looks at the prompt. She thinks about the pause before question three that she never documented in the certified record because she could not certify it.
She opens a documentation session. RECOL flags it as a voluntary supplementary annotation, which means it will sit in her personal archive under the session record, unweighted, possibly never consulted again.
She types: 1.2 second pause. Before question three. The question had not yet finished.
She closes the session. The lights are still dim. The proposal is open on a different window.
She does not know if the new argument is good enough. She will know when she sends it out for pre-validation, when the Board technical reviewers come back with their questions, when someone who read Torres v. RECOL more carefully than she did finds the next catch.
What she knows tonight is that she used the right word in 2028 and she will have to write around it for the rest of her practice. This is not a problem to solve. It is a condition to work inside of.
The brief is still ninety-four pages. Page 34 still says potentially.
She makes a note in her observation log: Correct and certifiable are not the same thing. I have been trying to turn correctness into certifiability. The question may be different: can a practitioner's presence certify the conditions that produce correct observation? Begin there.
At 2:03 AM, she closes the archive. The RECOL interface returns to standby. The certification district is quiet outside the window — no extraction suites running at this hour, the overnight quiet the Board mandates under the session ethics framework. The lights are dim. She has been awake for three hours longer than she intended.
She does not feel like she solved anything. She feels like she finally read the brief.