Attest → the story
One visit. Three accounts.
Nothing rewritten.
The Alvarez residence. A scheduled morning care visit — Maria Chen is the worker booked to do it. What follows is not a dramatization — every timestamp, count, and quoted sentence below was read out of the same signed pack you can download. Each link verifies its source in your browser, no install, nothing leaving the page.
— the pack's own verdict, self-issued and re-checkable by anyone
(hover the shield for the whole verdict line).
The plan
Morning care visit booked for Maria Chen — 90 minutes at the Alvarez residence. The schedule is an expectation: it tells the record what to watch for, never what happened.
What the house saw
The household's own devices — doorbell, contact sensor, camera — reported 9 events during the window; the remaining 1 entry is the worker's own self-report through her scoped link, signed like everything else (all times UTC):
- 6:31 PM arrival motion webhook
- 6:31 PM snapshot local_or_test sha-256 pinned
- 6:31 PM doorbell webhook
- 6:31 PM door opened webhook
- 6:31 PM door closed webhook
- 7:36 PM door opened webhook
- 7:36 PM checkin unknown
- 7:36 PM door closed webhook
- 7:36 PM departure motion webhook
- 7:36 PM snapshot local_or_test sha-256 pinned
Each entry carries its ingestion source (webhook
= signed delivery; local_or_test = device snapshot, bytes
attested by a signed sha-256 and withheld from this redacted pack;
unknown = the worker's self-report — a person's claim, not a
device's). Coverage: observed with events — the receipt bounds its own window
(6:31 PM – 8:06 PM UTC) and signs 32% of it as polled, saying so in its own
words: "Attest polled Ring Event History during this window; it attests what the pipeline observed, not what physically happened" — so silence where it occurs is watched
silence, still never proof of absence.
The record closes — and says what it can't say
rcpt_8af58f6242d04… ·
verify this receipt →Door activity spans 65.25 minutes of the 90 scheduled — self reported, departure unconfirmed, identity never verified. The receipt self-labels its own limits:
departure_unconfirmed— Record closed for review; direction, identity, and continuous presence are not establishedobserved_interval_short— Observations span 65 min; 90 min scheduled. Time worked is unknown
signature_scope: record_integrity_only
— the signature attests the record's integrity, nothing about physical
truth.The humans spoke — 4 statements, all signed, none edited
Then the humans disagreed. Every statement below appended to the same append-only review chain, verbatim, under its own signature. The record doesn't merge them, rank them, or pick a winner:
“I dispute this record — I arrived before the first observation shown.”
“My mother was home all morning and says the doorbell never rang — nobody came to the door before ten.”
“I was home that afternoon � I heard the doorbell a little after six and saw someone at the door. I can't say how long she stayed.”
“I was home that afternoon; I heard the doorbell a little after six and saw someone at the door. I can't say how long she stayed. (Re-sending — my earlier message had a garbled character.)”
Read them again: the household's account contradicts itself — first "nobody came," then "I heard the doorbell a little after six." That contradiction is the evidence. A system that normalized it away would be destroying the only thing the family actually has. And the pack carries doubt the same way: records signed inside a revoked key's suspect window stay annotated on the green stamp — never hidden, never rewritten.
Then somebody changed a number
A second copy of this pack exists with exactly one signed field rewritten:
observed_span_minutes inflated 65.25 → 135.25 — past the
90-minute schedule. It still looks identical: same files,
same story, same layout.
Open it in the
verifier → FAILED: payload hash mismatch. Or
compare the two packs
— the page verifies each on its own merits, then names
observed_span_minutes as the moved field and proves the
transparency log only appended between exports. And
this is the page a coordinator
hands a mediator — attest diff --html renders the
same comparison as an attachable report.