When an LLM writes your tests, it tends to write tests that confirm what the code does rather than tests that verify what the user experiences. Your test suite goes green. The app still breaks in ways none of those tests would catch. Marlene Mhangami from Microsoft makes the case for flipping the order: get the agent to write failing Playwright tests against the expected behavior first, then generate code to pass them. The demo runs this live with GitHub Copilot and the Playwright MCP server on a toy store ...