Why you shouldn't vibe code without tests

A practical walkthrough of adding end-to-end tests to a vibe-coded application using Playwright and Cursor to prevent regressions. Covers writing a happy-path test for a video generation app, mocking third-party services (ElevenLabs, Fal.AI, OpenAI) to avoid costs, setting up repeatable test baselines, using plan mode with powerful models like Claude Opus, and automating test runs via GitHub Actions on pull requests. Emphasizes that even non-coders doing AI-assisted development need tests to catch breaking changes before they reach production.