AI
Cursor
The editor I reach for when a change spans many files.
- Score
- 4.0
- Status
- Still using
- Reviewed
What I tested
Six months of day-to-day work on a production codebase: refactors that touch dozens of files, test backfills, and the unglamorous work of wiring a feature through layers.
What worked
- Multi-file edits are where it earns its place. A rename that spans the route layer, the API layer and the tests is one prompt and a review pass.
- Knowing the open file and the recent edits makes suggestions far more relevant than a chat window that starts cold every time.
- It stays an editor. The familiar keybindings and extensions still work, which is why it survived the novelty phase.
What did not
- It will confidently produce a diff that looks right and is not. If you cannot review the change, you should not ship it.
- Context selection still needs babysitting on a large repository. Too much context is as bad as none.
Verdict
The best fit for engineers who already know the codebase and want leverage on mechanical work. It amplifies judgement; it does not replace it.