Credibility registry
Claims, evidence, and limits.
A claim, its source, and what that source can establish. Reviewed September 5, 2026. Private results below remain owner-reported; a diagram, code excerpt, or demo is not independent verification of a customer outcome.
| Claim | Definition and period | Available evidence | Limits |
|---|---|---|---|
| ~$10K written premium influenced | Owner-reported pilot result from policies closed by two producers. Premium is not commission, agency revenue, or software revenue. Measurement window not published. | Private policy records; attribution model described in Referral Desk. | Public aggregate, policy count, and confirmed/probable/assumed breakdown are not supplied. |
| 14 public helper tests · Referral Desk | Two runnable modules: header-secret authentication and duplicate decisions. | Public CI and source. | Does not run the private application, database tests, tenant integration excerpt, or provider-signature tests. |
| 1,666 private checks · Referral Desk | Private suite reported on 2026-08-30; 92 unit / 37 integration / 3 end-to-end files reported. | Owner-reported private test run. | A current report tied to the full revision and deployment gate is not published; public CI verifies a separate subset. |
| 93 tests · Load Desk | Test count reported on 2026-08-29. | Owner-reported node:test run; sandbox. | The demo demonstrates a workflow, not a test run. A revision-linked test artifact is not published. |
| 2–3 weeks → same day · Load Desk | Workflow design comparison: outsourced Manual-J versus the proposed in-app field process. | Synthetic-data sandbox. | Not a measured field result. No measured time saving or productivity percentage is claimed. |
| 88 checkpoints · ProWAX | Owner-reported deployment coverage for the staffing agency whose crew works an Uber facility. Deployment date not published. | Private system; walkthrough on request. | Checkpoint count is scope, not a measured productivity gain. A dated, permission-safe deployment artifact is not published. |
| Approved for use in a Mountain View school district. | Classroom Narrative AI approval is owner-reported; approval evidence is held privately. Public Guest demo is available; school deployment status is not established here. | Architecture and demo. | Product-source comments are not independent corroboration of district approval. Permission-safe approval evidence is not published. |
| 336 checks · Classroom Narrative AI | Private test count reported as 336 checks across 26 files. | Owner-reported unit, end-to-end, and accessibility suites. | Current run date, revision, and downloadable test report are not supplied. This page does not claim a new test run. |
| $300K+ device sales · résumé | Device-sales total; résumé describes roughly $30K per month for nearly a year. Exact reporting window not published. | Résumé; underlying sales records private. | Not software revenue; supporting aggregate or employer corroboration not published. |
| $22K premium / $150K–$220K quoted pipeline · résumé | New written premium in month one and quoted pipeline; distinct measures, neither is agency revenue. Exact window not published. | Résumé; underlying producer records private. | A redacted period aggregate is not published. Do not add these amounts to Referral Desk’s influenced total without checking overlap. |
| $15K / 10 website engagements · résumé | Web revenue and engagement count for JDG Marketing. Recognition basis and measurement period not published. | Résumé; underlying client and payment records private. | Booked, invoiced, and collected amounts require separate reconciliation before a more specific claim. |
| Lead/opportunity count withheld | No count is published until the unit, definition, period, and attribution basis are reconciled. | Private pilot records. | No inferred denominator or conversion rate is presented. |