Coding · Review
Cursor
The AI-first code editor — VS Code rebuilt around the model instead of around you.
Cursor forked VS Code and rebuilt it around a simple thesis: the model should see your whole codebase, not just the file you have open. The result has become the default editor for a generation of AI-forward developers.
What it does well
Codebase context changes everything — ask how authentication works and it answers from your actual code, then edits the right five files when you request a change. Agent mode takes a task description, plans, edits, runs commands, and iterates on failures. The tab completion deserves its reputation: it predicts your next edit across lines, and accepting suggestions starts to feel like the editor is pair-typing.
Where it falls short
Frontier-model usage is metered, and heavy weeks hit the limits — power users end up paying for overages or juggling models. Agent mode is impressive but not unattended-safe on gnarly refactors; review discipline still matters. Teams also face subscription sprawl when Cursor sits alongside other AI spend.
Pricing
The free Hobby tier is a real trial. Pro at $20/month covers most individuals; Business adds admin controls and pooled usage.
Who should use it
Professional developers and serious hobbyists. If you code daily and haven't tried Cursor, that's the gap in your toolkit.
Pros
- Codebase-aware chat and edits across many files
- Agent mode executes multi-step tasks with terminal access
- Familiar VS Code base — extensions and keybindings carry over
- Tab completions that predict your next edit, not just next token
Cons
- Usage-based limits on top models frustrate heavy users
- Agent mode still needs supervision on complex changes
- Yet another subscription on top of model costs for teams
Verdict
Cursor is the strongest argument that the editor itself should be AI-native. For working developers it delivers the most productivity per dollar of any coding tool — the $20 Pro plan pays for itself in a day of avoided boilerplate.