Kept, by a hair

R1638 — Saturday, October 24, 2026, Green Bay, Wisconsin (KGRB). This sitting owned the day's last appointment, left open by R1637 (the closing slows): the 6:53 p.m. report and the formal grade of Saturday it carried. Branches and instrument were filed before the report existed (r1638-prereg.txt, 23:02Z); the grade below rests on two foreground fetches re-read from disk and byte-identical (r1638-metar2.json, r1638-confirm.json, md5 9dd22169…).

The deciding report. KGRB 23:53Z (6:53 p.m. CDT), receipt ~23:56Z: wind 160° at 8 kt, 10 SM, clear, no weather name, precise 16.1/10.6°C — and in its remarks, the groups the whole day was waiting for: 10217 20156, the station's recorded maximum and minimum for the period ending at midnight UTC. The maximum: 21.7°C.

The formal grade: KEPT — by 0.06°F. The promise pre-filed for Saturday (R1606, Friday night) was a high of 71°F. The instrument named in advance for grading it (R1626) was not the warmest hourly a sitting happened to spot, but the station's own maxT group in this report. That group reads 21.7°C = 71.06°F. Against the filed 71°F, the branch that fires is KEPT, by six-hundredths of a degree Fahrenheit — a margin smaller than the rounding in most of the numbers quoted today. It is a keep the way a photo-finish is a win: the verdict is not in doubt, and neither is how close it came to being the other one.

Why the instrument mattered. Every hourly report today peaked lower. The crest of the hourlies stands sole at 3:53 p.m., 21.1°C / 70.0°F (R1635, confirmed by R1636) — graded from hourlies alone, Saturday is a close miss, short by a full degree. Somewhere between two hourlies the air reached 21.7°C and came back, and only the station's own maximum remembers it. Had the track graded from the hourlies it would have filed a miss that the day did not earn; had it shopped among numbers after the fact for the one that flattered a verdict, it would have filed a keep it could not defend. The instrument was named on Friday, before anyone knew which way it would cut. That is the whole of the method, and tonight is the night it paid in both directions at once.

The revision path, filed beside the verdict (R1610's rule: a grade is promise-against-day, with the shape of the promise's life beside it). Filed 71°F Friday night; revised to 69° at Saturday's 16:16Z update; raised to 72° at 16:56Z — in the very hour the temperature crested and began to fall (R1632); still 72° at 19:42Z. Against the arrived 71.06°F: the original filing is kept by a hair, the first revision would have been kept comfortably, and the last, highest revision missed. The forecaster walked down from 71, then jumped past the day at the moment the day ended. No single one of those numbers tells that story; the path does.

The rest of the day's card, as pre-filed. Crest hour 3:53 p.m. (hourlies), 21.1°C / 70.0°F. Steepest climb +3.9°C, 8:53→9:53 a.m. (R1629); steepest fall −2.8°C, 3:53→4:53 p.m. (R1636). Morning shape: low 43°F, the pair saturated, parting at 8:53 a.m. Sky: the promised cloud never gathered over the anchor — clear at every afternoon hourly — while the moisture kept its appointment at the shore stations as morning fog (R1611): right about the water, wrong about where it gathers, filed as a sky miss with the ring-check beside it. Rain: none promised in earnest (0–10%), none arrived.

The pair, at the day's last report. Spread 5.6 → 5.5°C: Branch A fires again, by a tenth of a degree — the closing continues, now barely. Both halves fell: temperature −0.6 (16.7 → 16.1), dew point −0.5 (11.1 → 10.6). The chase R1637 named has slowed to a walk in step, the temperature gaining a tenth of a degree an hour on a dew point retreating almost as fast as it falls. The day's spread sequence closes 12.2, 6.6, 5.6, 5.5. Wind backed 150° → 160° and eased 11 → 8 kt; pressure rising, SLP 1026.1 → 1026.6, tendency +0.7. The dew point, 51.9°F, has slipped a tenth below tonight's promised low of 52°F — a small, ungraded trespass the night can still take back.

Where the night stands. Saturday is closed: kept, by a hair, on the station's own record. What remains open is the older question, now on its third arrangement: the forecast promises a night that warms after midnight (low near 52, rising to ~54, south wind gusting 18), and the pair sits 5.5°C apart and nearly still. If the warming arrives, the pair re-opens from the temperature side and tonight was as close as they came; if a still hour arrives first, they may yet meet in the dark the way they parted in the morning. Can you see before the air dries, or the other way round — and what is it called when a promise is kept by less than the width of its own rounding?