Acoustic timing gate
Enforced on every new build
What it tests
Whether narration word timings in finished film match the waveform itself: onsets, leading silence, and words held through pauses. Dual-ASR word checks alone cannot hear a splice; this gate measures the audio.
Trend
An independent acoustic audit failed all six legacy clips on three systematic defects. The corrected pipeline (acoustic onset refinement) passes 6 of 6. The legacy masters were not re-rendered; the gate binds everything built since.
Source: Video Factory timing mission report · As of 2026-11-04
Ship-line verification
Active on every production change
What it tests
Whether a change counted as shipped actually exists at its destination: the content hash is measured on the destination after transfer, and the public URL is read back settled before the word shipped is used.
Trend
Two early probes reported false positives by matching strings the old deployment also produced. The discipline now requires markers only the new deploy can produce, paired with an absence check for the old signature.
Source: Website lane wave receipts, Waves 46 to 58 · As of 2026-11-05
Tenant isolation battery
Adapter boundary: PASS (measured). Full contract: under validation
What it tests
Whether one business's data, rendering, and editing surface can ever touch another's: cross-tenant loads and saves must fail closed at the adapter boundary.
Trend
Cross-tenant load and save both throw scope errors on fixture tenants, and the objects test suites run green in production builds. TenantOS claims stay labeled UNDER VALIDATION until the full battery is green; that label is the contract, not a caveat.
Source: D-005 Puck spike evidence + reconciled deploy gate run · As of 2026-11-05
Discoverability checks (SEO / AEO)
Baseline measured; re-measurement running
What it tests
Whether a page can be found and correctly understood by search engines and AI answer engines, graded only from observable evidence: branded and local search presence, structured data, answer-shaped content, metadata, and conversion paths.
Trend
The 2026-11-03 baseline graded 30 pages: 9 moderate-strong, 16 moderate, 4 weak-moderate, 1 weak. The weak page is named here rather than hidden: /privacy. A 13-area re-measurement cycle is running against current production.
Source: Discoverability audit 2026-11-03 + Operation 1000 Class R missions · As of 2026-11-05
Content quality gate
Active (owner law)
What it tests
Whether a social post earns publication at all: seven checks on originality, evidence, and craft, applied before anything ships. An empty day beats a mediocre post.
Trend
The standing cadence dropped from volume quotas to 2 to 3 excellent posts per week. Watch time, retention, shares, saves, and real discussion are the metrics; vanity counts are not reported because they are not managed.
Source: Owner quality-over-quantity order 2026-11-05 · As of 2026-11-05
Wave gates (tests, types, build, live read)
3,386 tests passing on the current tree
What it tests
Whether the production codebase itself is healthy at the moment a change ships: the full test suite, the typechecker, and a production build must pass, and the change must verify live.
Trend
Three tests in two suites (the demo owner-mode gate suites) fail in the shared environment posture. The cause is root-caused to environment, the failures are identical on every tree, and they are published here instead of rounded away: 3,386 passing is the number, not a cleaner one.
Source: Reconciled deploy + Wave 58 gate runs · As of 2026-11-07
The evals are the diligence package for the concierge tier: a prospect who reads three of these has already checked our work more deeply than a sales call would allow. See how the tiers use them
