Ask HN: Anyone used LLM for UI E2E testing?

Did it work? Was it better than writing cucumber (or any other feature) tests?