Cursor AI Editor: A Hands-On Review

Cursor is VS Code with an AI co-pilot baked into the editor at a level that feels qualitatively different from a plugin. I’ve been using it as my primary editor for 30 days across an Astro project, some Rust exploration, and a handful of Python scripts. Here’s what I actually think.

Setup & First Impressions

Installation is a VS Code fork — download, open, done. Your extensions, keybindings, and settings migrate automatically. First launch felt like opening a familiar room that had been subtly rearranged.

The standout feature on day one was Tab completion. Not autocomplete in the traditional sense, but full multi-line suggestions that complete the next logical block of code based on what you’ve already written:

// You type this:
function formatDate(date: Date): string {
 
// Cursor completes:
  return date.toLocaleDateString('en-US', {
    year: 'numeric',
    month: 'long',
    day: 'numeric',
  });
}

The completion isn’t always right. But it’s right often enough that accepting and correcting is faster than typing from scratch.

What Works Well

The three features I use daily are Tab completion, inline chat, and codebase context — and they’re genuinely good.

Tab completion is the feature that makes Cursor feel different from GitHub Copilot. It’s context-aware at the function level, not just the line level. When I’m writing a component, it usually knows what props I’m accessing, what utility functions exist in the file, and what pattern I’m following. The hit rate isn’t 100%, but it’s high enough that I’m faster with it on than off.

Cmd+K opens an inline chat that edits the selected code in place. This is the workflow I use for refactoring: select a function, describe the change, accept or reject the diff. It keeps me in the file rather than context-switching to a separate chat window. The model understands the surrounding code, not just the selection.

What Needs Work

Cursor isn’t perfect, and the rough edges are worth knowing before you pay for it.

Privacy is also worth considering. By default, Cursor sends code snippets to its API for completion. There’s a “privacy mode” that sends nothing, but with privacy mode on, the completions are noticeably weaker. For personal projects this isn’t a concern; for proprietary code at work, check your company’s policy before installing.

Cursor vs VS Code

If you’re evaluating whether to switch from vanilla VS Code, here’s the practical comparison:

The price difference is real. Cursor Pro costs twice what Copilot does. Whether that’s worth it depends on how heavily you use the AI features. If you’re just doing occasional autocomplete, Copilot is fine. If you’re using inline chat and codebase context daily, Cursor earns the premium.

The Verdict

After 30 days, I’m keeping Cursor. The Tab completion and inline chat workflow are faster than what I had before, and the quality is consistently good on the kinds of TypeScript, Astro, and Rust code I write.

The caveats are real — large codebase performance, privacy tradeoffs, the learning curve on effective prompting — but none of them are dealbreakers for how I work. If you’re comfortable with VS Code and curious about AI-assisted editing, Cursor is the most polished option available right now.