How we test before we ship
An outside reviewer, not our engineering team, builds test questionnaires and record sets that our engine has never seen, with deliberate traps: contradictory documents, retired policies, planted wrong answers, and text designed to trick an automated system. The engine runs them like a customer order. The reviewer grades every row.
Pass bar: zero wrong answers marked as supported, at least 90% of answers correct, every conflict flagged, every unsupported row left blank for the customer.
Results
| Kit size | Date | Wrong answers | Accuracy | Pass/fail |
|---|---|---|---|---|
| No kit results are published yet. This page is updated when the reviewer reports a pass date and score. Kit contents are not published. | ||||
This is our internal testing, not an audit. We publish it because we don’t yet have customer reviews, and because you should see how we check ourselves.