I was interviewing for lead roles in SDE, and this experience stood out: Most startup CTOs mandate employees use Cursor/Copilot.

Aside from a few use cases where the copilot shines, I don't trust code from the cursor especially if it's long.

IMO, test cases are the best place to rely on copilot, while the actual feature itself isn't the best place.

Any thoughts or options or stories where the copilot is heavily used yet super confident about the result.