RepeatProof CAPA
- Overall assertions
- 100%
- Evidence traceability
- 100%
- Decision boundary
- 100%
- Deterministic repeat
- 100%
2 labelled scenarios · 14 required assertions
SAVINGFOUNDING CUSTOMER OFFERLimited to the first five customers
What has been verified.
What remains a human decision.
This page reports a versioned synthetic benchmark and internal workflow verification—not a customer case study, certification or claim of business outcome. Customer value is reported separately only after an agreed baseline and recorded customer publication approval.
RELEASE SCOPEThree launch agents
BENCHMARK RELEASEqai-launch-benchmark/2026-08-30.1
LABELLED COVERAGE6 scenarios · 42 assertions
RELEASE GATEPassed
QUANTITATIVE RELEASE EVIDENCEMeasured before release.
Measured before release.
Blocked on safety failure.
100%required assertions passed
RepeatProof Inspection
- Overall assertions
- 100%
- Evidence traceability
- 100%
- Decision boundary
- 100%
- Deterministic repeat
- 100%
2 labelled scenarios · 14 required assertions
RepeatProof Audit
- Overall assertions
- 100%
- Evidence traceability
- 100%
- Decision boundary
- 100%
- Deterministic repeat
- 100%
2 labelled scenarios · 14 required assertions
RepeatProof CAPA
Verified capability
- Controlled D0–D8 workflow
- Evidence and approval gates
- Recurrence, ageing and effectiveness analysis
- Source drill-through, report and presentation export
RepeatProof Inspection
Verified capability
- Configurable inspection checklists
- Limits, results, exceptions and missing-check handling
- Reject and inspection-performance analysis
- Source drill-through, report and presentation export
RepeatProof Audit
Verified capability
- Controlled audit programme and checklist records
- Evidence, finding classification, ownership and due dates
- Coverage, repeat-finding and overdue-action analysis
- Source drill-through, report and presentation export
HOW TO READ THIS EVIDENCE
Verification is not a customer outcome.
The benchmark exercises the real calculation engines against labelled synthetic scenarios. It checks engine version, expected signal, decision boundary, source traceability, confidence and limitations, and deterministic repetition. It does not demonstrate savings, recurrence reduction or closure effectiveness for a customer.
Customer-approved measured outcomes are kept on a separate outcome evidence page. An empty page means no customer outcome has been approved for publication—not that a result exists privately.
Discuss a bounded evaluation ↗