← AR Lab

Shift handover

One shift, from both records: what the instruments crossed, and what a person wrote down. Rung 11 generates a note from telemetry and it is a good note and half a note — it knows the bearing crossed 4.5 mm/s and it does not know somebody put a hand near it forty minutes earlier.

What it is looking for

Where a person's note and an instrument's crossing are about the same machine and close enough in time to be the same event, the gap between them is the finding — and its direction decides what it means.
Person after instrument: the alarm was already up; the note confirms it.
Person before instrument: somebody knew first, and the system took that many minutes to agree. That is the whole argument, and it is invisible unless both records sit on one clock.
A note is reported against both the crossing it followed and the one it preceded. Picking whichever tells the better story would be choosing the finding in advance.
A shift with no observations says so out loud. An unrecorded shift and a quiet shift must not read the same.

Which half is which. The instrument events are simulated — the same twelve-hour night rung 11 replays, mapped onto the last twelve real hours so both records share a clock. The observations are real: whatever you logged in rung12 on this phone. Nothing shifts an observation's timestamp; moving the only genuine data in the room to make it line up would be falsifying it.

node tools/check_handover.mjs — 13 assertions, including that a note between two crossings reports both, and that an observation from outside the window is excluded rather than dragged in.