QA warns: translating requirements 3x is the real problem, not test count
trash_lover_79 · reddit · 2026-09-08
A QA engineer argues the core waste in testing is translating the same requirement three times: Jira ticket → handwritten test cases → Playwright/Selenium automation, with every requirement change requiring updates in all three places.
He's unimpressed by AI that dumps generic cases like "validate button works"; instead he wants requirements to stay the source context while AI generates candidate cases that QA can accept/reject/fix before they become automation. He highlights KaneAI by TestMu AI as an interesting fit — it takes Jira/PRD context, generates scenarios, allows refinement, and converts useful ones into executable tests (Jira integration still in Beta).
Longer term he wants one business flow connected across UI, API, and DB rather than three independently green tests. He stresses garbage-in-garbage-out: QA still must catch missing negative cases and weak acceptance criteria, and asks whether anyone has actually run this requirements → cases → maintained-automation loop without accumulating hundreds of unwanted tests.
More from coding & agent
- Codex CLI tip: Ctrl-O copies the last response as Markdown, /copy grabs just the relevant part — gabrielchua · 2026-09-08
- herdr 0.9.0 ships the most-wanted feature: one client to control agents across all your machines — ycombinator · 2026-09-08
- Hands-on with Grok Bot's cloud agent team: hiring an AI staff without a server closet — RachelVT42 · 2026-09-08
- Should Agent Skills scripts be prewritten or generated on the fly? — dotey · 2026-09-08
- Embedflow: zero-downtime embedding model migration without 108-day backfills — Potential_Low_1183 · 2026-09-08
- agent-browser now records browser agents at 60fps with --fps 60 — HankYeomans · 2026-09-08