The Hours That Didn't Feel Wasted

There was a time when I used to code a lot. Really, a lot.

I made around 3.5k contributions to GitHub repositories in 2021 and around 4k in 2022. Looking at those numbers now feels a little unbelievable to me.

The pandemic had something to do with it. At least, I can find one positive side to it :) While many of my peers suddenly had a lot more time to prepare for their boards, competitive exams and everything that comes with them — no school, no time wasted commuting — I was one idiot who got completely caught up in my love for AI.

It all started with the Intel AI4Youth program. CBSE had partnered with Intel to build an AI student community and I somehow became very involved in it.

I had always been interested in computational thinking. At South Point, we were introduced to QBasic, Visual Basic and a few other programming tools (forgot now) and I remember liking the idea of making a computer do something through code. When we started learning Python in school, it never really went much beyond the board curriculum for me. I didn’t understand what high-scale projects you could actually build with it. AI4Youth changed that for me.

There were workshops, bootcamps, projects and more importantly, a lot of people who were there simply because they wanted to learn. I became one of the more active contributors in the community. For the first time, I had a national platform where I could interact and discuss things with people from all over the country who shared the same interest. I had always been interested in STEM awareness and that community gave me a place to actually do something about it.

I started an initiative there. I helped people, organized coding contests and eventually got more involved in the community. Later, another student, Krish (a very talented junior who, if I remember correctly, was selected for ISEF) and I received the best initiative award. That motivated me even more. I was made a student representative.

I started building an online e-learning platform to conduct activities for the community. I remember waking up, opening my laptop and immediately starting to work on some new feature. Online classes were going on in the background. I’d just go downstairs for lunch or dinner and come back to the laptop. Anyone looking at my schedule probably would have assumed I was spending all my time preparing for school and exams. I wasn’t. I was coding. That curiosity definitely came at a cost academically then, but I also have a lot to thank it for. It was probably one of the most important things that got me genuinely more interested in ML and shaped the direction I eventually took.

There were no AI coding tools back then. If something didn’t work, I had to figure out why. Bugs meant reading documentation, searching on StackOverflow - a lifesaver (you might love Jaidev’s blog), digging through old forum posts, trying things, breaking something else, fixing that and slowly understanding what was actually happening. And this is where all those GitHub contributions came from. And when I look back now, I’m honestly surprised that I managed to build a reasonably decent version of an LMS back then without AI assistance at all.

Fast forward a few years and AI has completely changed the way we write code. Back then, writing code was the only way I knew how to build things. Now, it is possible to build things without writing most of the code myself. This year, I’ve gotten back into webdev while building a new version of a quiz management system with advanced proctoring features for the final version of our event at Paradox.

And something strange has happened. I now sometimes feel blank while writing or understanding fairly basic stuff. The language isn’t new to me. I’ve worked with it before when there was no AI. But I haven’t used it seriously in a long time. And now I have AI. I can explain exactly what I want, describe the logic, specify the edge cases, tell it how different parts of the system should interact and get a huge chunk of code back in secs!! And honestly, it is incredibly useful.

But there’s a catch. Finishing a coding task doesn’t always feel as satisfying as it used to. Earlier, when I finally got something working after spending hours debugging it, there was a certain excitement to it. I knew exactly how much effort had gone into making those few lines work. Now, I can give AI a detailed prompt describing exactly what I want and get a working implementation back. The logic can be mine. But sometimes I find myself looking at the syntax and thinking, “Wait, how exactly does this work again?” That part, I didn’t really expect. AI didn’t make me worse at thinking. In many ways, it has made me faster at getting from an idea to an implementation. But because I don’t have to write every line anymore, I don’t always have to remember every line either. And that matters. I miss that little sense of accomplishment that came from figuring it out myself.

There is a difference between being able to describe what a program should do and being able to sit down..implement it yourself from scratch. I think I have become much better at the first one and.. in some areas, noticeably worse at the second. That’s probably the cost of what people now call “vibe coding”. You describe what you want. AI generates it. You read through it, make some changes, run it, fix a few things and move on. For building things quickly, it’s fantastic. For maintaining your programming fundamentals, maybe not so much.

The irony is that I used to spend hours debugging code because I had no other option. Now I can ask an AI to explain the error, rewrite code or generate the whole thing again.

Thank God we have AI now…

But maybe I also need to thank the old version of me who spent those hours debugging everything manually. Because I’m starting to realize that those hours weren’t just wasted time. They were practice. If I stop writing code myself, eventually I’ll have to ask it how the code works. And that’s probably a cost worth remembering.

(PS. Some random collection of thoughts I penned down while staying up late with my team, coordinating the event. Finally, making it online now with the new version of my website, almost 1.5 years later. Btw, this post is not written by an AI system.)