Introduction

Since the advent of ChatGPT and other AI tools, I can’t recall the last time I developed a fully functional app without relying on them. And I suspect many developers reading this feel the same way.

It’s been two years since GPT tools revolutionised development workflows, and the pace of progress has been staggering. As AI continues to advance, the temptation to rely entirely on these tools grows stronger. At first glance, it might seem like delegating coding tasks to AI saves time and effort, but there’s a hidden cost: the gradual erosion of your own skills and expertise.

In this article, I’ll share how I fell into the trap of over-relying on AI, how I broke free, and how I now use these tools to up skill and become a better developer every day.

The Problem

The main problem of GPT and other AI tools is the easiness and how good they are. You can just type in “Create an API to fetch list of products” and it will give you production ready code. You just need to copy and paste it and it will work. Heck nowadays we have AI in built in the code editor. That just makes it easier to slack off.

Let’s walkthrough using an example. Let’s say you want to build a todo mobile app. Now that everything is so easy, you just write a prompt “I want to build a todo app. Help me writing code for the homepage”.

You get the code, you paste it, you run and everything is working fine. You think alright good results, now let’s move on to second page, third page and so on. Suddenly, the AI hallucinates in one of the responses and you encounter a bug. Since you have written 0 lines of code, you are definately not going to debug this.

Next step? Copy the buggy code, paste it in the chat box and hope that the AI gives you the solution. In a lot of cases it does, but when it doesn’t it just doesn’t. You go on and on with it chatting and trying different prompts, giving it context of n number of files in the hope that it will solve it for you.

And when you lose all your patience, that is when you thing of debugging it for yourself and search the error on google and try to find a solution. In my experience, most of the times the problem was that the buggy code was outdated and was using functions that the latest version of framework/library/package didn’t support. Such a simple and common problem and you wasted so much time with gpt.

You didn’t waste time because you are not skilled to debug, you did it because you were lazy. Since you wrote 0 lines of code, you never did put in the hard work and you were never really active. Hence, you just never felt like actively debugging the code and find the solution. You think that oh i already know how to code in this language or framework, why should I write this let the AI write the code I already understand what the code does.

I mean debugging is an art in itself and getting your code to work after trying various methods after reading the docs or understanding what actually is wrong gives an immense satisfaction. At least it does to me.

So yeah, this is the big problem. We end up relying on GPT from minute 0 and in that process, we waste our time. Also, majority of the times when this happen, we end up learning nothing from the project or the work that we are doing.

The Solution

Now that we have gone over the problem, let’s look at how I changed the way I code with AI’s help.

First of all, I stopped writing prompts such as “Write code for this page” or “write an api”. I started looking AI as an expert reviewer instead of an expert creator.

Going back to the same example, what I do now is I write the code for the home page of the todo app. I then select particular parts of code, paste it in the chat box and ask GPT to review it. I then also go through the review and actually try to understand what it is trying to tell. Because, it might be that the review is not helpful so can’t blindly copy paste AI’s code again.

In this way, I also remember the changes GPT suggested and i learn new things as they are small changes made by AI’s review. Also, if i face a bug, I know what code I have written and since I am actively writing code, I debug the code without showing any laziness.

I also noticed that while writing code, i rarely paste my error output in GPT first. I go to the google search bar first and if I don’t find any help from Stack Overflow or Github issues, then I turn to GPT.

This helps in saving time. This also helps me up skill as I learn from my mistakes rather than making no mistakes at all and just seeing AI write my code. This also gives me the happiness of debugging my own issues.

This keeps me confident in my skills and makes me believe that no this AI won’t take my job as it is not the creator but the reviewer.

Conclusion

At the end of the day, the most rewarding part of being a developer is solving challenges, building something meaningful, and growing through the process.

When you sit to write code, you think a lot. You think of how you will place the div, how you will structure your functions, how you will create the interfaces and much more. When you make AI write code from scratch, you miss doing all this.

AI can be a powerful tool to help you along the way, but your unique abilities—your critical thinking, creativity, and determination—are what truly make you irreplaceable.

I would love to hear your views on this topic. On that note. Goodbye and Keep Coding!