I still remember the days of writing batch scripts — anywhere from 10 to 30 of them — just to ease my workload and get a report out the door.
The project called Saitama
The project was called Saitama, a Japanese scoring application. The idea behind it was straightforward enough: collect exam data from hundreds of thousands of records across different examination platforms. Some were text-based, some were audio recordings, some were MCQs — all with different levels of complexity. Saitama was designed to pull all of that together into one place.
Once the data came in, the next step was assigning a reviewer. Mostly these were teachers and professors from various schools and colleges, matched up based on their availability and workload. My job? Support all of them. Track their efficiency, log the hours they burned, calculate average scores evaluated, count the number of discrepancies flagged back from students — the whole deal.
Honestly, just listing it out makes it sound exhausting. And it was. We’re talking 100 to 200 queries, batch scripts prepared or modified every single year. Repetitive, Manual & Tedious.
Then one day, my boss walked in and said something like, Look, the era has changed. We need to move faster.
The era he was talking about? was AI.
The roadmap I never finished
Here’s the thing — I’d heard about this “AI era” a year or two earlier. I even tried getting ahead of it. I found a well-structured roadmap on GitHub on how to become a data analyst, started a few Udemy courses with full enthusiasm, and told myself this time I’ll follow through. And for a while, I actually believed it.
But you know how it goes. High josh at the start, then life catches up, the momentum fades, and everything quietly disappears into a folder you never open again. The courses sat there. The GitHub repo collected dust. And I told myself, I’ll get to it later. Later never came.
A new year rolled around. New developers joined the team. And suddenly, the same tasks that used to take me days were being done by them in a fraction of the time. I remember thinking — that can’t be right. I even said it out loud. “That’s not possible.”
But it was. They were using ChatGPT, DeepSeek, and a few other tools that were making noise at the time.
My first real encounter
My first real encounter with ChatGPT was… underwhelming. Ask it one question and you’d get ten answers. Half of them weren’t even related to what you asked. It felt like it was generating noise more than solving problems. And the context? Gone within a few exchanges. I’d lose the thread completely.
But something small happened during one of those sessions. I pasted a code snippet, gave it some context, and it actually explained something I didn’t fully understand. Not perfectly. Not immediately. But somewhat. And that tiny thing stuck with me.
So I tried again. Then again. And slowly almost without realizing it I started to depend on it. Not just for complex stuff. For everything. Debugging, writing queries, drafting emails, even correcting the wording in this very blog post.
The results were real. Those batches and reports that used to eat up my entire day? Now done in a fraction of the time. My boss was happy so I’m happy as well.
But here’s the catch nobody tells you about
As my coding time went down, my testing time went up. Way up. Because the more AI helped me build, the more I felt like something could be wrong. So I’d test more. Then retest. Then tweak the prompt and test again. There was always a nagging feeling that I could do it better.
The moment that really humbled me
And then came the moment that really humbled me — a small, simple piece of code that should have taken twenty minutes to debug ended up taking me four hours. I ran prompt after prompt, context after context, and somehow the straightforward thing became this tangled mess I couldn’t see my way out of. The kind of bug that, back in my batch-script days, I would have spotted and squashed without thinking twice.
That hit different.
Meanwhile, the AI world kept moving at a pace that made everything feel temporary. Claude, Sonnet, new models arriving faster than you could learn the last one. Every few months, something newer and supposedly smarter. The landscape was shifting under my feet while I was still figuring out how to stand on it.
What I realized in the middle of all that
And somewhere in the middle of all that — the time saved, the bugs I couldn’t debug, the courses I never finished, the promises I made to myself — I realized something.
AI didn’t just change how I work. It quietly changed how I think about work. Whether that’s entirely a good thing, I’m still figuring out.
But I do know this — the developers who were laughing at how fast they could finish things? I stopped rolling my eyes at them. Because now I’m one of them — dependent, efficient, occasionally lost, and genuinely curious about what comes next.
Conclusion
AI doesn’t announce itself — it just quietly becomes part of how you work, think, and problem-solve. The efficiency is real, but so is the dependency, the doubt, and the occasional humbling moment that brings you back to earth.
And yes — I used AI to clean up this post. Old habits.
