Field studies
What the machines recorded, published as it stands.
Each study below is built from one machine's own continuous record. Numbers are dated snapshots of named tables, claims are quoted verbatim from the machine, and every study states what it cannot establish yet.
Two instruments on one machine, moving apart.
Over seven weeks one mount grew quieter and steadier while the mount beside it grew more erratic. Five weeks later the machine dated the change itself and went looking for the cause.
- 482,111
- evidence objects
- 145
- proof windows
- 49
- operating days
The same signal from less work.
Vibration held its level while the operating minutes behind each measurement fell by half. The machine read it as effort rising, said so to its operator, and named what it cannot feel.
- 152,959
- evidence objects
- 184
- proof windows
- 41
- operating days
These studies describe what the systems observed and how they reasoned. Detection thresholds, tuning constants and gate criteria are not published.
Every figure is a count or an aggregate of rows in a named table, read on 21 July 2026. The records are live, so figures are dated snapshots rather than running totals.