The day I forgot how to code.
I’ve always loved backend development, especially building things with JavaScript. During my internships, I went deep into it and had a blast.
Then, in early 2024, GPT-4 and Claude Sonnet 3.5 dropped and completely blew me away. I thought, these tools are incredible—why wouldn’t I use them?
But I made a mistake: I started treating them like a crutch instead of just tools.
Whenever I needed to implement a new feature or fix a bug, I’d copy the console error, paste my code into a prompt, and let the AI handle it. Fast. Easy. Done.
Two minutes and voilà—problem solved.
My productivity went through the roof. I was shipping a feature a day, moving at breakneck speed. It was addictive.
But that speed came with a hidden cost.
Ever heard of technical debt?
Technical debt is what happens when you keep patching problems instead of fixing them properly. And I was racking up debt fast.
The site started crashing randomly. Unverified users were somehow getting access tokens. That’s when it hit me: my codebase was a disaster.
The AI couldn’t save me anymore. The context window couldn’t keep up with my spaghetti code, and the only person left to untangle it was me.
What I found was ugly: redundant code everywhere, logic that made zero sense, hooks used in bizarre ways. It looked like five different people wrote it—but it was just me, copy-pasting AI fixes without ever understanding them.
AI hadn’t failed me. I had made myself dependent.
The struggle to fix things on my own.
I knew which module was broken, but when I sat down to rewrite it, my brain froze. I’d lost the ability to think through the whole workflow.
Every roadblock triggered the same reflex: ugh, this is annoying, I’ll just ask ChatGPT. It had become automatic—and that scared me.
So I made a decision: no AI until I could think in code again.
As a student, my job is to learn and improve. By leaning too hard on AI, I was robbing myself of the very reason I was coding in the first place.
Learning to code again.
It was rough at first. My hands literally stalled at the keyboard. Even writing a simple for loop felt like climbing a mountain.
But I stuck with it. Now, I force myself to write an initial solution before I touch AI. I’ll still brainstorm with it, but only after I’ve thought things through myself.
And the result? I’m a much better programmer now. The struggle is where the real growth happens.
A small piece of advice.
If you’re learning too, don’t let AI take away the thinking part of programming. These tools are powerful—but if you skip the hard parts, you’ll pay for it later.
Because once you forget how to think in code, trust me—it’s a scary place to be. And it’s completely avoidable if you use AI with the right balance.