state
The snapshot is derived, never authored. It is regenerated wholesale from the event log on every build, so a hand edit does not survive — it is caught because log_head would no longer match a recomputed chain.
Snapshot
- schema_version
- 0.1.0
- mode
- live
- generated_at
- 2026-08-16 00:52:38Z
- build_ref
- f64abec3046d52227d4c0a8875c611f23bc26bee
- log_head.seq
- 303
- log_head.hash
- e9edd34fe267fbaff3b13aff3357ec2758d14e315b42944fe5ef4346c413aef1
Autonomy
- events_total
- 303
- automated
- 271
- seed_bootstrap
- 32
- human_intervention
- 0
- ratio
- 0
- last_human
- —
producer_triggers
- human
- 5
- scheduled
- 13
- system_event
- 1
- not_machine_produced
- 3
Human intervention is not part of the normal publication path. These figures come from the log, so they are checkable rather than asserted. An event recorded as human_intervention must also carry a reason.
Counts
| claims | 7 |
|---|---|
| evidence | 19 |
| experiments | 12 |
| falsifications | 3 |
| limitations | 26 |
| external | 0 |
| challenges | 5 |
| events | 303 |
| external (withheld) | 1 |
Counts are of published records. Records held back by the publication boundary are reported as a withheld figure on each collection endpoint, so that absence stays distinguishable from omission.
Checking this yourself
Every event carries prev_hash and a hash taken over its own canonical serialisation with hash removed. To verify that nothing was rewritten between two visits, fetch the events endpoint, walk the chain from seq 1, and compare the head against the value you recorded last time. A rewritten record anywhere in the log changes every hash after it. /api/events
This is the property that makes the surface worth reading: you do not have to trust the summary on this page, because you can recompute it.