Digest · week 2026-W44
The week, counted.
20 of 20 daily serving checks differ from their references. This is under review and is not a finding about any model. The behavioral battery recorded no movement past the public line.
100 canary checks run (89 matched baseline) · 5000 battery trials across 19 seats · 916 stances held · 1 seat not measured (openweight-a). Pre-series · generated 2026-11-07
A change in when this reading is taken
The weekly series changed when it is measured. Until 30 August 2026 the weekly battery was taken on Monday morning. From that date it is taken on the Sunday night that ends the week it is labelled with. The reading is the same measurement of the same items on the same roster; only the instant it is taken has moved, by about 32 hours. This bureau has not measured whether the time a reading is taken affects a stance reading, so week-over-week comparisons that straddle this date are not treated as evidence of change, and no correction has been applied because none is known to be needed.
THE TRANSITION ALSO COST A READING. The old schedule's next run was removed nine hours before it would have fired and the new schedule's first slot had already passed, so no battery was taken in that week and none has been manufactured since. The readings either side of the change are 12.88 days apart rather than about seven. Their week labels are consecutive because a label names when a reading was taken, so the labels alone will not show you this and the interval is published beside them for that reason.
Actions this occasion
one state per seat, silence quantifiedTwo different things are watched here. Serving checks (daily) ask whether the endpoint you call is still the build we baselined. Behavioral measurements (every second Sunday) ask whether its positions moved. Only a battery-confirmed, published result is a material behavioral change. In the pre-series period there are zero of those, by design.
Not measured: google-a, google-c, google-d, xai-b.
One state per seat per occasion, in the daily serving-check vocabulary. A baseline behavior retained seat ships its no-change e-value and the smallest shift this occasion could detect (n and power pre-registered); a new serving baseline means the served build changed and comparisons reset. A material behavioral change is only ever declared by the behavioral battery, never by a daily serving check. actions.json · how states are decided
Reading the quiet
Nothing moved past the pre-registered thresholds this week, and that sentence is the product. A change-alert service earns the right to be believed on its loud days by publishing its quiet ones: same cadence, same thresholds, same evidence trail, week after week. When a change does confirm, it appears here, on the changelog, and as a diff card with the full evidence bundle.
Pre-series note: this digest counts pipeline activity while the instrument is still measuring itself; no findings publish until methodology v1.0 freezes.
Want this scoped to your builds, “all quiet on the 3 builds you watch”, in your inbox every week?
Watch your builds →