How Cursor Changed My Relationship with AI

For years, I’ve jumped between editors — Brackets for its clean UI and live preview, Sublime Text for speed, and VS Code for its all-rounder versatility. But a few months ago I switched to Cursor, and it’s completely changed how I write code. It feels like a true AI assistant.

Remember Sublime and Brackets?

Both had their charm. Sublime was super-fast and minimal. It was perfect for when you just wanted to get in, fix something, and leave without a single popup or plugin slowing you down. Plus, they had great themes which I would switch between every now and then, keeping it feeling fresh.

Brackets, on the other hand, was the designer’s editor: clean, web-focused, and ahead of its time with live CSS editing.

But both felt static. They were great text editors — not coding partners.

Enter Cursor

Cursor feels like what Sublime would have become if it had evolved alongside modern AI tooling. It’s built on top of VS Code, so you get all the familiar shortcuts and ecosystem — but layered with something far more intuitive: AI that actually understands your context.

You can literally talk to it about your code.

Not just, “what does this error mean?” but, “Refactor this Next.js API route to stream responses instead of returning a single payload.”

And it actually does it, neatly, respecting your project structure, variable names, and even commenting the logic as if another developer had stepped in to help. The more I've used it, the more I feel like it understands my style of coding and replicates it as if I would have done it. Insane.

Why It’s Revolutionary for Developers

For me, the biggest shift has been moving from manual recall to contextual reasoning.

In Sublime or Brackets, you had to remember everything — syntax, imports, utilities, and every past workaround ever used (or copied from StackOverflow).

With Cursor, you just describe what you’re trying to achieve and let it come up with the blueprint. It remembers the file you’re in, understands related imports, and can reason across files — something traditional autocomplete could never do.

When I’m building something complex (like a Next.js feature pulling from Payload CMS, or setting up edge function logic for Vercel), the AI isn’t just finishing my sentences — it’s completing my thought process. Often before I've even thought that far ahead.

Why It's Not Replacing Me... Just Yet

In the early days, I would have an existential crisis every other day. One day, I'm excited by the possibilities of it all and how much time it will save me. Then next, I'm convinced it will take over and render me redundant.

The more I've used Cursor, the more I've realised it's great. But no way near perfect.

Often it will run into errors, cause mistakes it hasn't accounted for. Sometimes the way it tackles problems is just bizarre and complete overkill (often creating endless new 'test scripts' to run and make sure a feature is working as expected, never considering to clean them up).

It's even tried to install a non-existent package via NPM, completely adamant it does in fact exist (watching it fumble when it couldn't provide a link to it's documentation felt extremely vindicating).

The Bottom Line

The last big leap in text editors was probably when VS Code unified the dev experience across languages. Cursor feels like the next one.

It isn't perfect by any means. It definitely needs a developer to guide it and evaluate every change it proposes. The best use case is to explain what needs building and exactly how to get there. Then letting it do the bulk of the foundational work, ready to evaluate by a developer to dot the I's and cross the T's.

As someone who writes, designs, and prototypes daily across multiple stacks, it’s hard to overstate how much this has reshaped my workflow. I’m no longer “writing code” — I’m collaborating with it.