Alt Report

Findings written like a teammate, not a scanner.

Only Confirmed findings interrupt anyone. Each one arrives with a narrated replay, the evidence and a suggested fix.

#qe-releasesalt.qa · 16:04

Release 4.18, booking funnel: NO-GO. 60 cells, 34 passed.

What broke: step 4 “Continue” depends on a consent checkbox with no label, below the fold on phones. It stopped 14 cells across every agent: that makes it a product bug. Replay + fix

What changed since 4.17: 4 fixed (keyboard access to rooms, price summary, cookie wall, accented names). 2 new, including Gemini now getting a 403 on /checkout.

Fix first: the checkbox (P0). Ruled out: payment issues (every passing cell reached the payment page and stopped).

AI-generated summary · Claude Sonnet 5 · limitations apply

Narrated 60-90 s repro

First-person narration over the real recording, labelled AI-generated. The part no other QA tool ships by default.

Preview →

Confirmed-only tickets

Jira, Linear or GitHub, with replay, DOM, suggested fix and a repro command.

GO / NO-GO one-pager

Pass rates, top findings, model versions and the limitations statement. Board-ready.

Preview →

MCP for coding agents

Claude Code or Cursor fetch the failure bundle, fix it and re-run to verify.

Learn more →

The finding, then the one-pager for your VP.

Engineers get the step, the evidence and the fix. Leadership gets GO or NO-GO with the top findings. Product preview on sample data.

A one-page release readiness report marked NO-GO with the top findings.

A Confirmed finding with what happened, verification, and a suggested fix.

See where agents fail on your funnel.

Run one free cell on a property you own. No sales call, no card.

Run a free cell →Book a pilot