Evidence notes / September 2026
A useful pilot starts with evidence.
NoryBot is an early product. These notes separate the paths we have exercised from the questions your team should test on its own application.
Beyond the login screen.
On 9 September, NoryBot generated and ran 20 cases across two live applications: five Cypress cases, ten public Playwright cases, and five authenticated Playwright journeys. The protected journeys exercised search, product configuration, selection dialogs, saved results, and navigation back to the same result.
| Scope | Runner | Passed |
|---|---|---|
| Public content and navigation | Cypress | 5 |
| Public sign-in controls and pricing | Playwright | 10 |
| Authenticated product workflows | Playwright | 3 |
| Authenticated saved-result journeys | Playwright | 2 |
Both protected suites needed one automatic locator repair before passing. Their actions and expected outcomes stayed fixed; the generated files were not hand-edited. Removing the session made all three product tests fail. Deliberately enabling a disabled control caused the expected assertion to fail, and restoring the application returned the same suite to three passing tests.
These are dated, supervised observations. They do not establish whole-application coverage, SSO/MFA support, or reliability on every application. Generation and repairs can take several minutes and consume provider usage.
Security and installation checks.
The full release verification on 28 September passed authentication, generated-test execution, security regressions, tenant and role matrices, browser proof, evidence reports, and fresh customer ZIP installation. Browser proof uses real Chromium against controlled fixtures; generated security contracts are also exercised through the supported test runners.
Fresh installation and same-version reinstall were checked with Cypress and Playwright, including preservation of project data. These checks do not substitute for a customer's first provider call, paid activation, or a future database migration.
Generation through execution.
On 7 September, an internal two-page browser fixture generated Cypress and Playwright suites using the Abliteration provider. The generated files ran without manual test edits: two Cypress tests and six Playwright tests passed.
The fixture covers a small form, greeting updates, and navigation. These results show an end-to-end path, not a production-app success rate. Generation took several minutes.
Billing and access.
Sandbox acceptance covered purchase, license visibility, activation, paid scans, the two-device limit, deactivation, and cancellation behavior. No real charge was made. Renewal, refunds, and email delivery remain outside that acceptance run.
What to establish in your pilot.
01
Choose one representative flow with a clear expected outcome.
02
Check the generated selectors and assertions against your requirements.
03
Run the suite, introduce a known regression, and confirm the test catches it.
04
Track the editing and maintenance effort your team actually spends.
05
If evaluating Pro, inspect scan scope, findings, and incomplete coverage.