Cursor

AI · Code editor

The editor I use to move quickly through unfamiliar code, refactors, debugging, and agent-assisted implementation.

Why it earns a place

Why I use it

The useful part is not autocomplete. It is being able to reason over a repository, ask for targeted changes, compare the diff, and keep the human decision loop close to the code.

Use cases

Best for

Exploring a codebase before changing it
Small scoped refactors with visible diffs
Generating tests or migration plans
Pairing with Git so every AI change remains reviewable

Start small

Quick start

01
Open the repository, not isolated files
Give the editor the actual project so references, types, routes and conventions can be inspected together.
02
Ask for a plan before a wide change
For anything cross-cutting, first ask which files are affected and what can break. Then implement in small checkpoints rather than accepting a giant opaque patch.
03
Review the diff
Treat generated code like a pull request from another engineer: read it, run the app, test edge cases, and revert aggressively when complexity increases.

Repeatable workflows

Things I come back for

Bug hunting

Paste the exact error, ask the agent to trace the execution path, and make it identify the smallest reproducible cause before touching unrelated files.

1
Reproduce the error and capture the exact message or broken behavior.
2
Ask Cursor to trace the smallest relevant execution path.
3
Apply one narrow fix, then rerun the reproduction before accepting broader cleanup.

UI iteration

Give the agent the design constraint and the current component, then ask it to remove complexity before adding another abstraction.

1
State the visual and responsive constraint in plain language.
2
Ask for the smallest component/CSS change that satisfies it.
3
Inspect the result at mobile, tablet, laptop, and wide desktop before merging.

Documentation

After a working change, ask for a concise README update that explains the decision rather than narrating every line of code.

1
Finish and verify the change first.
2
Ask for a short explanation of the decision, inputs, and operational gotchas.
3
Remove generated prose that merely restates the code.

Reference

Links