How I Got Too Dependent on AI (and How I’m Fixing It)

Hey devs,

Welcome back to Chiristo’s blab corner. Today I want to talk about a topic that’s been on my mind for a while AI and Vibe Coding. Honestly, this is something I feel every junior engineer (including me) is struggling with. AI is powerful, no doubt. But here’s the truth

Too much of anything is good for nothing.

Back in my first year of college, I was introduced to the magical word programming. I came from a bio-maths background, so everything felt new to me. On top of that, starting college during the corona period meant I missed the induction programs that usually give students a sense of how domains like computer science fit into the real world.

In my first year, I had Python in my syllabus. That’s when I started learning Python and programming fundamentals on my own. Slowly, I got into the flow. By my second year, I became kind of a nerd. A few of us teamed up for a state-level hackathon, where the challenge was to build an advanced chatbot for government schemes. Ours was the only team that made it through from our college, which got us hyped. But when it came to actually building the solution, reality hit. We just didn’t have the skills yet and ended up failing badly. The reason? We hadn’t even learned Software Engineering and we lacked the skill set and programming knowledge.

Still, I kept improving myself in different domains web development, mobile development, and more. I mostly learned through Google and YouTube. Then came the game-changer ChatGPT. I used it to speed up my projects. In my third year, me and friends participated in a national-level hackathon and even won an award. That time, we applied our critical thinking and used ChatGPT to enhance our solution.

This is where things started going off track. I slowly slipped into vibe coding. Earlier, I typed every single line of code and understood what I was doing. But later, I started using ChatGPT even for tasks that required critical thinking. I was depending more on AI than on myself.

During my internship, I had to build a full-stack chat application with React as the frontend and Node.js as the backend. The right approach would have been to begin with system architecture. But what I actually did was run to ChatGPT, grab a template, copy-paste it, and when errors showed up, I copied my whole code back in and asked it to fix it. I was just copying whatever it gave me without really understanding anything.

That’s how I got trapped. Initially, I used ChatGPT for things I already knew, just to save time. Over time, I started leaning on it for stuff I didn’t even know yet. That was the first time I am using Context API, Redis, and a few other concepts. Instead of learning them, I just let ChatGPT hand me answers.

By the end, ChatGPT and I had completed the project together. But during the code review, I realized I couldn’t even explain my own code. I had barely typed 10 lines per file (and most of them were just logging statements). I had to ask ChatGPT to explain the code it generated. That hit me hard.

I told myself:

I was never a real coder, let alone an engineer. Yet AI made it seem like I was.

It struck my ego. I am an engineer and a coder, but not solving problems on my own felt completely wrong.

That’s when I decided to transform myself into a true developer, not just someone who vibes with code. So, I created a roadmap in my calendar called Developer Transformation. Every day, I’ve set reminders for practicing DSA to improve problem-solving, learning system design and cloud fundamentals to understand how real-world systems work, and doing coding sessions without AI to build confidence in my own skills.

One big thing I figured out was learning to use AI the right way. I came across a blog from my senior Use AI Like a Supplement, Not a Steroid and it really made me stop and think. I realized what went wrong not just with me, but with a lot of junior devs. AI should be a supplement to my transformation, not the main source of it.

In these 1.2 years of working, I’ve watched how seniors deal with problems. They never rush to copy-paste stuff straight away. They break problems down, reason about them, and then build. That’s the skill I want to develop. And yes, AI can still be part of my journey, but only to accelerate it not replace it.

To be a real developer, you’ve got to solve problems with your own brain. AI can help, but at the end of the day it’s just a tool not the actual skill.

learn first, struggle first, then use AI smartly. That’s how I’ll grow into a real engineer.

This is the path I’ve chosen.

This is my developer transformation.