I joined a project with an existing Playwright E2E test suite, 38 spec files, ~165 tests, around 14,000 lines of test infrastructure. My first step was simple: run the tests locally.

8 out of 130 non-skipped tests passed. A 6% pass rate.

The confusing part? CI was green. It turned out CI ran everything with workers: 1 , multiple workers plus the dev environment meant running tests locally just wasn't possible.

Step 1: Analysis — asking questions I didn't know the answers to

I had zero domain knowledge of this codebase. No context on why tests were written a certain way, what the custom wrappers did, or where the real problems were. So I started asking AI to analyze everything, the Playwright configs, the page objects, the spec files, the CI workflows.…