Dormant Baseline
It is 10:13 PM and Marcus Veil is sitting at his desk reading a standards proposal he has already read.
The Branch 9 draft is beside him — three pages, the "junction that hasn't decided what it is" framing on the first page, the organizing logic on the second, the incomplete third. He is not going to read it tonight. Tomorrow morning he will read it from the beginning, fresh, at 8 AM when his eyes work correctly. He made this decision Tuesday afternoon and he is not undeciding it.
What he keeps reading instead is the CIRCM-7 proposal. Specifically section 4.2, "Historical Observations on Dormant-Baseline Calibration," third paragraph, footnote 11.
"See Comment 0047 (Veil M., 2024-11-18), submitted during the 90-day comment window for CIRCM-5 Dormant-Baseline Standards Review. Veil argues that the 45-day observation window is insufficient to distinguish true dormancy from false-negative dormancy in substrates with seasonal compaction variability. No response action was taken during the CIRCM-5 cycle, but the committee noted the methodological concern for future cycles."
No response action was taken.
Marcus is not sure what he expected to feel reading that sentence, but whatever he expected, this is not it. He feels the specific flatness of being right in a document. Not vindicated — vindication requires that someone was watching when you were right. This is something else. Comment 0047 sits in the CIRCM-5 archive in perpetuity with that number as its permanent identifier, and two years later someone pulled the archive, found the comment, found it worth citing. The comment has now done more work in a footnote than it did when Marcus originally submitted it.
He submitted it at 11:54 PM on November 18th, 2024. He remembers this because the 90-day comment window closed at midnight and he had been revising the last paragraph since 9:30. He hit submit with six minutes left and then sat in the dark for a while before going to bed.
The Branch 7 monitoring cycle was active at the time. The sensor array at Junction 7-C had been producing anomalous low-frequency readings for eleven days, and Marcus had a hypothesis about the drainage subgrade but no way to confirm it without a core sample that nobody had budgeted for. The dormant-baseline detection algorithm — the one running continuous 24-hour cycle checks on all monitored substrates — had logged the Junction 7-C readings as within-threshold on each of the eleven nights, classifying them as normal variation. Marcus had been watching the trend line drift in a direction the algorithm wasn't calibrated to consider problematic.
He put the hypothesis in the comment. That was not the smart move. A comment on a standards review is not a place for an unverified hypothesis — it is a place for documented patterns, cited observations, methodological concerns grounded in evidence. Marcus knew this when he wrote it. He wrote it anyway, buried in the third paragraph, hedged enough to not be embarrassing: In at least one monitored substrate (Branch 7, Junction 7-C), sustained low-frequency anomalies consistent with subgrade variability have persisted for periods exceeding 30 days without triggering the current dormancy protocol's detection threshold.
He did not write: I think the ground is moving slowly in a way the algorithm cannot see.
He wrote the hedged version and submitted with six minutes to spare and went to bed.
The footnote does not cite the hypothesis. The footnote cites the methodological concern — the observation period question, the distinction between true dormancy and false-negative dormancy. Someone read the comment, found that part useful, and clipped it for the argument they were already making. The unnamed junction, the specific number of days, the word "insufficient" — these traveled forward two years into a new standards review. The word "insufficient" is doing work right now in a proposal that will shape monitoring protocol for every Circuit Miles successor project in the regional network.
Marcus's name is not in the footnote. His comment number is.
He is not sure how he feels about this either.
The drawer is to his right. The Branch 7 sensor log is inside — not the processed output, the raw data. Sixteen months of sensor readings from the full Branch 7 array, transferred to a portable drive when the Circuit Mile was officially decommissioned and the monitoring system was taken offline. He transferred it because the data would have been deleted when the system went dark, and Marcus was not ready to let seventeen months of low-frequency anomaly readings disappear into the standard archive protocol where no one would look at them. He put the drive in the drawer and thought: I'll look at this properly when there's time.
He has not looked at it.
What he remembers is the shape of it. Junction 7-C ran the low-frequency anomalies from early November 2024 through late January 2025 — eleven weeks, two apparent peaks, then a subsidence into normal-range variation. His read at the time was seasonal: temperature differential causing expansion-contraction cycles in the drainage subgrade, which the sensor array was picking up as micro-displacement events too small to be flagged by the dormancy algorithm but consistent enough to show as a pattern if you were watching the trend line.
He was watching the trend line. That was his job.
The dormant-baseline algorithm was not designed to watch trend lines. It was designed to compare each nightly reading against a calibrated baseline value and flag anything that exceeded the threshold by more than a defined tolerance. The algorithm was doing exactly what it was built to do. Marcus's argument in Comment 0047 was not that the algorithm was wrong — it was that the calibration window of 45 days was too short to establish a valid baseline for substrates with seasonal variability. If the algorithm calibrated its baseline during the stable-period summer months and then ran monitoring checks through winter, it would classify seasonal compaction movements as within-threshold because the baseline didn't account for them. The algorithm would say: dormant. The ground would be doing something the algorithm could not see.
No response action was taken.
The CIRCM-7 proposal is responding now, two years later. The committee is proposing to extend the calibration window to 90 days and add a seasonal-correction factor for substrates in climatic zones with greater than 15% annual temperature differential. These are exactly the changes Marcus argued for. His comment, cited in footnote 11, is being used as evidence that the concern predates the committee's current consideration — that someone in the monitoring community was watching this problem before the committee got around to naming it.
If Marcus files a comment on CIRCM-7, he would attach the Branch 7 sensor log as primary evidence. Not to make the same argument again — the committee is already making the argument. To add: here is what a false-negative dormancy profile looks like in raw sensor data. Here is the specific shape of the thing the algorithm was missing. The argument is already in the proposal. The data would be more useful than the argument.
He decides not to open the drawer tonight.
This is not avoidance. The data will still exist in November. The CIRCM-7 comment window closes December 15th — 84 days from now, by his count. He marked November 15th as his decision checkpoint and he is not moving that date because he spent a Tuesday night in late September reading footnote 11 repeatedly. The drawer holds the evidence for a decision that does not need to be made for eight more weeks.
He reads the footnote one more time.
No response action was taken during the CIRCM-5 cycle, but the committee noted the methodological concern for future cycles.
The CIRCM-7 automated submission system is running its nightly processing cycle right now — 10 PM batch, indexing the day's new public comments, updating the running tally. Marcus Veil shows zero submissions for the current review period. This is accurate. He has not filed a comment. The December 15th window is still open. The tally will remain at zero until he decides otherwise.
In the CIRCM-5 archive, Comment 0047 exists at timestamp 2024-11-18T23:54:17Z, status: archived, download count: 4. Someone downloaded it four times. Marcus doesn't know if that's one person reading it four times or four different people reading it once each. He doesn't have access to the download logs. He knows only that the download count moved from zero to four sometime between November 2024 and now, and that the current standards proposal cites his comment number in a footnote.
The Branch 9 draft is still on the desk: three pages, the "junction that hasn't decided what it is" framing on page one. Tomorrow morning he'll read it from the beginning and decide whether that framing is the core argument of the chapter or a metaphor to cut. He said he'd give it fresh eyes at 8 AM. He has not broken that decision all day.
He turns off the desk lamp.
The sensor log is in the drawer. The comment is in the archive. The proposal is in a 90-day review window. The Junction 7-C anomaly ran for eleven weeks and then stopped, and Marcus logged it and wrote the comment and went to bed, and the ground under Branch 7 is doing whatever the ground does now without him watching.
That is not nothing. That is what monitoring is: being present for what the instruments record, and putting it somewhere it can be found later. Comment 0047 was findable. Someone found it. The work held.
He goes to sleep.
The CIRCM-7 window will still be open in November. The Branch 7 log will still be in the drawer. Some things hold by staying where they are.