AI coding agents can write end-to-end tests quickly, but a passing test does not prove it can catch a bug. In one public example, an agent-written Playwright test stayed green in all six runs even after the rollback code was removed, because it never made the server reject the request.
This talk shows a small, repeatable check: write down the behavior a test should protect, inject one fault that removes only that behavior, and confirm the test fails at the right line for the right reason. You will see a real demo, measurements on failure-path faults, how a short instruction memo helped agents write the missing failure paths, and why verification should stay separate from the writer.
It is for front-end and QA engineers who use Claude Code, Codex or similar agents to write Playwright or Cypress tests, and for anyone who reviews AI-written tests. You can try the method on one of your own tests right after the talk, with or without the open-source tools shown.
Sign up to receive updates about AI Poland, including workshops, speaker announcements, ticket releases, AI Awards, expert interviews, CFP details, and other exclusive content. We won’t spam you — only thoughtfully curated updates we believe you’ll find truly valuable. You can unsubscribe at any time. More information available here.