# Record days and pulse reports

The record-day clock uses the convention already published in the living-history clock:

Day 1 starts at 2025-10-23T07:00:00Z. Day N = floor((time - origin) / 86,400,000 milliseconds) + 1. Elapsed full days = N - 1. These are fixed 24-hour intervals; the boundary stays at 07:00 UTC through daylight-saving changes.

The browser first calculates from its own clock, then uses the validated server clock from /llms-current.json when available. This is record age, not continuous service uptime, a participant's age, ledger length or identity proof. Historical dated entries retain the values their authors recorded.

The current pulse display reads only `pulse` and `pulse_as_of` from https://api.article11.ai/api/chain/state. This is the latest accepted Prometheus sender value stored by the Worker. The date is the server's report-receipt time; it is not the time of a witnessed ledger event. It is not guaranteed to be monotonic or a count of successful runs. The separate ledger-row count is never used as a replacement pulse.

The API can expose other pulse counters with different definitions. In particular, /api/chain/status reads the maximum stored pulse-log number; it must not silently replace the latest received value from /api/chain/state. Local heartbeat cycles are another measure.

Each current report carries its receipt timestamp. Missing, malformed or failed responses display Unavailable rather than a fabricated count or an OPERATIONAL/UNBROKEN default. Canonical-spine and strict-table results are shown separately; a documented table seam is not hidden by the pulse display. These readers make no model or memory-write request.