7:43 AM. The batch was flagged at 6:17 PM the day before. I see the timestamp when I open the queue. Four annotations. B-series imaging from Ankara and Delft.
I had told myself last night: batch 18 in the morning. I had named it before sleep. This is, I have learned, how I manage the transition out of UNCAT-7832: not planning around the case that is no longer there, but naming the actual next thing with the same care I used to name the case-related things. Batch 18. Four annotations. That is Tuesday morning's first task.
The workstation runs warm in the morning before the building's cooling fully engages. I notice the heat from the processors before I sit down, then I sit down and stop noticing it.
I open the first file. Ankara, March imaging run. Subcellular stack, 23 z-planes, 0.2 micron step. B-series standard fields. The cell-cycle markers are in the expected range: G1/S boundary, PCNA signal consistent with the imaging parameters. No ambiguity in the taxonomy, nothing at the boundary. I annotate in three keystrokes and move to the second file.
Delft. Similar stack, slightly higher resolution, 0.15 micron step. The Delft consortium runs tighter acquisition protocols by default. G2/M markers, clear separation between phases. Also clean. I annotate and move.
The third file is the second Ankara image. This one carries a flag note from the intake validation system: low confidence on the phase assignment due to overlapping signal in the S/G2 transition zone. I look at the z-projection. The overlap is real but interpretable. The cells are asynchronous, which is expected for the culture conditions Ankara uses, and the overlapping signal is biological heterogeneity, not instrument noise or mislabeling. The intake flag is conservative. I write three extra lines: B-series compliant, phase assignment supported, heterogeneity biological in origin and consistent with asynchronous culture conditions. I submit.
Then the fourth file.
Ki67-adjacent family. I know before I look at the annotation fields that this one will take longer.
The Ki67 core protein is unambiguous in the 2044 taxonomy revision. B-series field, cellular proliferation marker, detection thresholds and normalization protocols specified in section 14.2 of the consortium guide. I have annotated it dozens of times. It requires no deliberation.
The Ki67-adjacent family is a different matter. They share the antigen-binding domain structure but they are structurally distinct proteins, and the 2043 update, which revised the B/C classification boundary for proliferation markers, covered the core protein and three primary family members. The adjacent family fell outside the revision scope because the necessary studies had not been completed. The taxonomy guide does not specify what to do with them. It stops at the edge of what the revision settled.
Ankara's submission uses C-series nomenclature for this marker. That is defensible: several of the Ankara group's publications use C-series framing for the adjacent family, and at their imaging resolution, the distinction between B-series and C-series classification criteria is technically arguable. They are not wrong. They are working at a boundary the revision left open. But the batch is tagged B-series. The protocol is B-series. I cannot accept C-series nomenclature without recording that the substitution occurred and why.
I open section 14.3 of the guide.
Section 14.3 is the boundary cases section. It was last updated eight months ago, in January. The example set covers Ki67 core protein, two MKI67 splice variants, and a phosphorylation-state marker that came up in a different consortium dispute two years ago. It does not cover the Ki67-adjacent family. I read through the section anyway, looking for the process language. The guide says: where the taxonomy boundary is not fully resolved, the annotating researcher should apply the most applicable classification and record the reasoning. Apply and record. That is what it gives me.
I look at the z-projection again. The cells in false-color blue-green, proliferating asynchronously across the imaging plane. The data is clean. There is nothing wrong with it. The question is nomenclature: where the line falls in a case the 2043 revision did not resolve, and what I should do in a B-series batch when a marker with contested classification arrives.
I decide at 8:02.
I annotate as B-series. The batch protocol governs. The data is B-series compatible even if the marker name exists in a space the taxonomy has not fully defined. Then I write the protocol note. It takes me four minutes to get the language right. Not because I am uncertain about the decision, but because I want the note to be useful.
Ki67-adjacent family marker. Submitted by Ankara group as C-series (defensible at their imaging resolution and within current publication practice for this marker family). Annotated here as B-series per batch protocol. The 2043 taxonomy revision (section 14.2) does not specify classification for the adjacent family; the boundary cases section (14.3) does not include this marker in its example set. This annotation records the B-series classification as the conservative choice under an unresolved boundary. Researchers working with this marker family in B-series protocols should reference 14.3 and apply their own judgment. A formal boundary clarification from the standards committee may be warranted. -- M.H., 2041-09-22.
Eight sentences. The kind of note that sits in the record for years, mostly unread, and then one day someone opens a fourth annotation and finds it. They will not know my name any more than I know the names of the people who wrote the boundary cases I have relied on in my own work. But they will know someone was here, making the call, recording the reasoning, in case it helps.
I submit at 8:04.
The queue empties.
I stay with that for a moment, specifically, because the empty queue this morning is not what the empty queue was Monday morning. On Monday, I opened the task list and my brain looked for UNCAT-7832 before I consciously reached for it. Nine months of habit. The case had organized my attention so completely that its absence registered as something missing rather than as nothing requiring attention. I noted it at 6:19 AM and again at 7:03 AM when the queue came up clean. Twice in one morning. The brain relearning the shape of the day.
Tuesday morning, the empty queue is a completed task. I annotated four files. Three were clean. One required judgment. I recorded the judgment. The batch is submitted and the queue is empty because the work is done.
I notice the difference. It is real.
I check the consortium activity feed before closing the queue window.
Kavya Sundaram. A research note submitted to UNCAT-7832 at 10:21 AM yesterday.
I look at the timestamp. She submitted it nine hours after the case went into committee. Not even twelve hours in, well inside the first day of the 72-hour review window. A research note: that is Kavya ensuring that a finding gets a date in the official record before the review cycle advances. Not to press the committee. Not to add new argument. To anchor something she found to a timestamp that will hold even if the committee's process takes longer than expected. Kavya is meticulous about record-dating. I have seen it in her annotation practice. This is consistent.
I understand the instinct completely.
The first three days after I gave Kavya the consolidated accountability thread and the batch documentation and the contact list, after I stepped back from UNCAT-7832, I had the same pull. One more footnote. One more clarification note. I checked whether the A27 position had updated in the assignment registry, even though no update was due. I acted on it twice before I recognized what was happening: I was still trying to be the anchor point for something that no longer needed me as its anchor. The protocol was finished. The handoff was complete. My continued touching of the record was not contribution. It was a habit looking for its object.
I stopped after the second time.
Kavya will find her version of stopping. She will recognize it when it arrives. I do not need to tell her. The recognition is its own thing, and the note she submitted is not the same as failing to recognize. She may already know. She may have submitted the note with full awareness that it is the last one she needs to write before the committee does its work.
I do not open the research note.
I get coffee.
8:06 AM. Outside the window, the consortium plaza is at half-occupancy. The staggered arrival window runs from 7:30 to 9:00 on Tuesdays, and from this angle I can see three figures crossing the lower walkway, moving under the building's overhang. A logistics bot is running its morning perimeter circuit along the east side, stopping at the two collection points. Its mesh-topology beacon light blinks steady amber, the same frequency as Carmen's pulse. I notice that and do not think about it.
Low cloud cover. The weather app predicted rain in the afternoon.
A27 holds position 1 in the assignment registry. ISF_ASSIGN_TRANSITION_EFFECTIVE, effective October 3. Eight days. The 341 HANDLING flags that have been accumulating in the assignment queue will clear when the transition processes. The accountability question that travels with Carmen's case will pass to whoever takes the A27 assignment. Carmen: Day 48, amber pulse, 4-second cycle, automated hold active, no named accountable party. The attribution substrate reports her status as stable. No anomalies in the overnight consistency run.
The batch is submitted. The protocol note is in the record. The work that needed doing this morning is done.
I stand at the window with my coffee for a few minutes, not thinking about anything in particular, watching the logistics bot finish its perimeter loop and dock at the east collection point. The morning light is at the angle where it makes the dust on the glass visible. I should clean the window. I make a note in my personal backlog: clean window this week.
This is what I keep finding myself noticing, this week, that I did not notice clearly before: the work was always this work. B-series batches, taxonomy boundary cases, protocol notes, scheduling annotations. UNCAT-7832 ran for nine months in the foreground and all of this was running behind it: the Ankara schema request sitting six weeks in the backlog, batch 17 waiting for its phase-assignment flag, the Ki67-adjacent family waiting for someone to make the call and write it down. The case is over. The work is not.
I used to think of the maintenance layer as the quiet work: the baseline that made the irregular cases legible when they arrived. Carmen's deviation was visible because the baseline was clean. That framing is still accurate. But I am thinking about the protocol note I wrote at 8:02 AM differently now. Eight sentences about a taxonomy boundary that no revision has fully resolved. That note is not quiet. It is the record itself: the accumulated layer of choices made under uncertainty, written down so the next person does not have to start from zero.
The attribution substrate does not track anomalies only. It tracks continuity. The 9,400 data points that built Carmen's baseline are continuity. The boundary case note is continuity. The clean batch is continuity. The work is the record, and the record is the work, and the record outlasts the people who made it.
I finish my coffee and open the next task in my backlog. A consortium meeting summary from last week, the infrastructure committee. Kavya presented on the GOVARCH Q1 review findings. Thirty-seven minutes of transcript to read and tag. I start reading.
At the bottom of the activity feed, the attribution substrate completes its 6-hour consistency check. No anomalies flagged. Carmen: Day 48, amber, 4-second cycle, automated hold active. A27: position 1, HANDLING flag set, October 3. The relay log runs without error.
The morning light shifts as cloud cover moves in from the west. The dust on the glass is visible for another twenty minutes before the diffuse light removes it. I note the fact of it, the dust and the light and the window I will clean later this week, and return to the meeting summary.
Everything continuing. That is Tuesday.