Baseline, diff, fix, silence: four statements from real sealed captures of our own site — not four consecutive weekly scans; each slot states its spacing.
Internal demonstration — artifacts from our own self-audits, not customer evidence.
Every digest on this page re-verifies at /verify, and each sealed bundle is downloadable unmodified.
What a monitored month says
Weekly monitoring records what changed on each monitored site.
This page shows the four statement formats that cycle produces, rendered from real sealed captures of our own site — not staged screenshots. Each slot states its true capture spacing, names its sealed bundles, and links the public verifier.
Week 1baseline
The baseline: a real self-audit of our own site, sealed with its evidence bundle — the initial-scan statement every later comparison is measured against.
True spacing: single capture, sealed 2026-10-01T16:26:12.489Z — the initial baseline scan statement
The diff: two sealed captures of this site with a real measured difference between them — the weekly-monitoring diff statement, computed by the engine, recomputable by anyone.
True spacing: two captures 2 days apart (sealed 2026-09-28T15:50:53.844Z and 2026-09-30T20:23:42.093Z), from our own sealed self-audit monitoring history — not consecutive weekly captures
Between the two captures, 3 of 15 measured content facts changed: H3 headings 10 → 17, lists 3 → 2, links to this site 40 → 41; rendered main-content words 1041 → 1260. Every field is recomputable by anyone from the two published sealed bundles.
Fix verified — honestly, not yet: no controlled FAIL → fix → PASS arc exists in our self-audit history, so this slot shows the receipt format a verified fix will use, from a real scan.
True spacing: single capture demonstration (the baseline capture, sealed 2026-10-01T16:26:12.489Z) — no fix arc claimed
Not a fix-verified arc — not yet. The block below is the receipt format, from a real scan.
We state it plainly: no fix arc exists in our self-audit history yet. Here is the receipt format — the sealed findings and remediation records of one real scan (our baseline self-audit capture), downloadable and re-verifiable at /verify. Our history does contain a same-day FAIL-to-PASS pair (sealed 2026-09-19T10:11:06.851Z and 2026-09-19T13:37:04.555Z), but the engine protocol set changed between the two scans (1.2.0 to 1.3.0) and the flip cannot be attributed to a controlled fix, so it is not presented as fix-verified.
Week 4silence
Silence is a deliverable: two sealed captures with zero measured change between them — the no-diff receipt, exactly as the engine sealed it.
These same comparisons render as the emailed weekly roster letter — a sample of it is below; what you read here is the receipt itself, sealed by the engine.
True spacing: two captures 28 seconds apart (sealed 2026-10-01T16:26:42.147Z and 2026-10-01T16:27:10.433Z), zero measured change — a statement about these two captures, not a claim about a quiet week
Between the two captures, 0 of 15 measured content facts changed: the same heading counts, the same lists, the same links, the same 1339 rendered main-content words. Anyone can recompute the delta from the two published sealed bundles and get the same all-zero record.
Sample — rendered from our own self-audit data, not a delivered letter
One Firm account, one monitored site — ours. The quoted transition is the real diff from Week 2 above: 3 of 15 measured content facts changed between the two sealed captures, and the classifier quotes its line verbatim.
The classifier line reads UNDETERMINABLE on purpose: both sealed bundles predate engine-build stamping, so site versus methodology cannot be separated without guessing. The letter states that instead — the same classification runs on every customer transition.
About this letter
This is your weekly roster letter. Once a week we compare every monitored site against its previous sealed scan and tell you exactly what changed — on your site or in our methodology.
When a site is listed as unchanged, that means our weekly comparison detected no changes on it. Byte-level churn with identical measured facts is never reported as a change.
Classification notice
Some comparisons this week could not be classified, so no change claim is made for them. This window is bounded — every future comparison is classifiable.
This week
1 monitored site recorded changes this week — every comparison is classified inside.
Changes since the last letter
- https://generativemetrics.com/
[UNDETERMINABLE] Classified as UNDETERMINABLE — engine_build stamps are missing on one or both bundles, so the methodology axis cannot be ruled in or out without guessing.
Classified 2026-10-01 16:27 UTC
Methodology
Comparisons run on protocol set gm-ai-consumer-compat/1.8.0, engine build 20d7a15cbbeaa42d874c9ca5f48a14700577d802c87a991024ebbb0ce872b2a3. When our methodology changes, this letter says so.
Changelog: https://generativemetrics.com/changelog
Errata: https://generativemetrics.com/errata
Silence log
No changes were detected on 0 of 1 sites this week.
Sites whose comparisons could not be classified are not counted as silent — see the disclosure above.
Next comparison window: 2026-10-05 06:00 UTC
Generated Mondays 06:00 UTC.
Period: 2026-09-28 15:50 UTC to 2026-10-01 16:27 UTC
Letter ID: 2026-W40
Sealed captures behind the quoted transition: sha256 cc15f2…c6f48 (cc15f22d32c1286a9318f1744ef86c06fb86bb65c3b5464cf76b04f7745c6f48) → sha256 a17de6…fa407 (a17de6676eb428950645d6ea7b5c50e2b977e711d7e3fb9e4cfe35156bafa407) — both re-verify at /verify.
These slots are built from real sealed self-audit artifacts with real timestamps: four captures of our own site made on 2026-10-01, three of which are published here (the silence pair was sealed 28 seconds apart), plus two captures from 2026-09-28 and 2026-09-30 loaded from our sealed self-audit monitoring history. These are not four consecutive weekly scans; every slot states its true spacing.
Our self-audit history also contains a same-day FAIL-to-PASS pair (sealed 2026-09-19T10:11:06.851Z and 2026-09-19T13:37:04.555Z, under protocol sets 1.2.0 and 1.3.0) and a transport-partial capture (2026-09-20T09:17:13.287Z, INCOMPLETE during an HTTP 403 period). Because the engine changed between the FAIL and PASS scans and the fix cannot be attributed to a controlled change, that arc is not presented as fix-verified anywhere on this page.
What the weekly version becomes
For agencies, the weekly version ships with Firm monitoring as the Weekly Roster Letter: every client's change statement, in your inbox every Monday — the sample above shows the letter's exact form.
Every report carries a methodology-currency receipt naming the methodology that produced it and whether it is current or superseded.
The same-day alert digest summarizes the day's monitored-site changes, capped at one email per day.
