JOEY FLORES
Published on

Why I switched to Neovim from Cursor

Authors
  • avatar
    Name
    Joey Flores

Cursor, the double-edged sword

If you're a developer, chances are you've sipped from the AI-code-generation fountain and was instantly hooked. I undoubtedly was.

It's beautiful... you think of some feature you want, enter it into a text box, press enter, and boom, wonderfully formatted code that mostly works. That first time you sit and think, "what kind of sorcery is this?" right before entering the the next desired feature. And then over and over, thoughts quickly turn into features.

Before long, you have a proof-of-concept application running in a fraction of the time it would've taken without Cursor. Yet, you have no idea how it actually works. If you're anything like me, you thought, "I'll surely take the time to understand the code being generated." We're good at bargaining against ourselves.

"Just one more request and I'll review this code. Wait, I need this too. Oh cool, it added this. Let's add this too."

39 helper, util, config files later... eh, commit and push. Congrats, a dangerous can of worms was just spilled across the floor.

Deliberate Friction

There's a certain resolve and apprecation for that very resolve that comes from doing hard things. As humans, we need difficulty not only to strengthen our character, but to accomplish an unfortunately declining human experience: learning.

ChatGPT May Be Eroding Critical Thinking, as stated from a TIME article reporting on an MIT study released in 2025 that found an accumulation of cogntive debt when using an AI assistant for writing an essay. Though studies in this domain are as novel as the introduction of LLM's to society is, the onset is troubling and worth noting.

So what is there to do with information?

For me, that meant reevaluating my growing usage (dependancy?) on LLM tools. Make no mistake, LLM's are in invaluable tool that will do immeasurable good to society. The ChatGPT input bar (I almost referred to it as a search bar, proving my next point) has all but entirely replaced my Google usage, as it has with many others. The classic, "Google it," has rapidly evolved to, "ChatGPT it." Google can't single-handedly allow me to train for multiple marathons all contained within a single chat thread, but ChatGPT can and will continue to do so for me. There are those information-seeking use cases for LLM's where it thrives, almost stealing the show for us humans when it comes information retreival.

Learning follows repeated information retreival. Snip that effort from our brains and outsource it to an LLM, well you simply don't learn. Therefore, despite the quick, shiny LLM-generated code Cursor is able to spit out, I would stare at it without much satisfaction. It looks like it mostly works, but why? A question that would take too long to answer sifting through the 39 helper, util, and config files.

I truly believe it's in our human nature to learn and progress to achieve any level of satisfication with one's self and belonging in the world. Cursor was limiting that for me. Neovim, in all its' tradition, swings that door open and invites me to slow down, create, and truly feel connected to my work.