My Experience as a Software Engineer in the Modern AI-Focused Development Era

This is an article I’ve been thinking about writing for quite some time—mainly about my experience working with AI and how it has affected the way I build software.

For people who don’t know me, I started my journey into software development back in my university days around 2016. Even before that, I was always interested in computers and hardware.

The Pre-AI Era

Back then, for university projects, we used the tools available to us at the time: good old debugging, reading documentation, Stack Overflow, or some random YouTube video with less than 100 views that somehow solved the exact issue you were facing.

After completing my degree, I joined the industry and continued working in pretty much the same way.

I also had a small quirk as a developer—I always tried to complete tasks and projects to the highest level possible. I cared deeply about architecture, clean implementations, maintainability, and doing things “the right way.” Sometimes that caused a few headaches for the teammates I worked with, but I always managed to deliver on time.

As I grew as a developer though, I realized that constantly chasing perfection isn’t always feasible. You have to find a balance between what’s ideal and what’s practical. Sometimes quick-and-dirty solutions are necessary to get things done.

That was me before the rise of ChatGPT, Claude, Gemini, and the modern AI wave.

My First AI-Assisted Project

I vividly remember the beginning of the AI revolution. At the time, I was still working at my first company and was tasked with developing an Android SDK.

That was the first real project where I used AI to actively assist my development process.

The project was a production-grade SDK, and I was working with Kotlin—a language I wasn’t very familiar with at the time. Naturally, I was a bit nervous about jumping straight into development. So I started using Claude and ChatGPT to help me design a solid architecture and guide parts of the implementation alongside the traditional methods I mentioned earlier.

And honestly, that project became one of the best projects I’ve worked on.

But even then, my workflow was still very primitive compared to today—basically copy-pasting between AI chats and my IDE. No integrated AI agents. No automation. Just conversations with ChatGPT and Claude alongside manual development.

The Shift at Oloodi

After that, my process became a mix of AI-assisted copy-pasting and traditional engineering techniques.

It wasn’t until I joined Oloodi that I fully stepped into AI-agentic development, and that completely changed my development patterns. Today, my workflow revolves heavily around AI agents—but not in the sense that I’ve handed over my entire job to them.

Not at all.

I see AI as a teammate with almost infinite potential.

I’m still constantly reviewing code, adding guardrails, validating decisions, and steering the overall direction. The AI helps accelerate execution, but the responsibility and engineering judgment still sit with me.

What AI has really allowed me to do is return to that “perfect as possible” style of development I used to enjoy. Before AI, there was always pressure to compromise—to take shortcuts, move quickly, and accept technical debt because of time constraints. But with AI agents, I can now focus much more on best practices, clean architecture, coding standards, proper patterns, and overall software quality without feeling like I’m constantly racing against the clock.

The Hidden Benefits

Another huge benefit is side projects.

Previously, balancing work and personal life made it difficult to consistently work on ideas outside my job. But with modern AI tools, building side projects has become significantly easier and honestly much more enjoyable.

AI has allowed me to spend more time on the part of software engineering I enjoy the most: planning systems and designing architecture.

Ironically, AI actually makes architecture even more important.

If your planning is poor, the output will reflect it. There are already a few personal projects I’ve completed using this workflow, and several more are on the way.

What I’ve Lost

Now, to talk a little about the downsides.

One thing I genuinely miss is the debugging journey—the satisfaction of spending hours tracking down a frustrating issue and finally fixing it. That feeling was special.

These days, AI often solves problems so quickly that you sometimes skip that entire experience.

But I’ve realized I now get a similar excitement from learning deeper concepts—understanding new technologies, better architectural patterns, scalability strategies, system design, and engineering principles that I previously didn’t always have time to properly explore.

The Risk for New Developers

That said, I do see a major risk for developers who are just entering the industry.

These AI agents are incredibly powerful, and sometimes they allow people to build things without truly understanding the fundamentals of software engineering or the technologies they’re working with.

And that’s dangerous.

Today, having a strong foundation is more important than ever.

I like to think of it this way:

Imagine you’re the head engineer building a house. The AI agents are your workers. They can build incredibly fast and execute almost anything you ask for—but if your understanding of the foundation is weak, the instructions you give will also be weak. And the final result will reflect that.

Bad prompts usually come from shallow understanding. Good engineering still matters. Probably more than ever before.

So don’t fall into the trap of relying on AI without understanding what’s happening underneath.

A Quick Note

A lot of how I think about AI comes from my day-to-day work at Oloodi.

We spend most of our time building AI-driven products and turning ideas into real, working systems. Being in that environment naturally changes your perspective.

At some point, AI stopped feeling like just a tool.

It started feeling more like working alongside another engineer.

And when you start seeing it that way:

• You let it challenge your ideas
• You expect it to check the latest documentation
• You rely on it—but still guide it carefully

These stop becoming “nice-to-have” behaviors and simply become part of how you build software.

If you’re working on something similar—or even just exploring the AI-first approach—feel free to connect or reach out. I’m always happy to exchange ideas and learn how others are approaching this shift.

And if this article helped even a little, feel free to share it with your team.

Thanks for reading—and happy coding.

Originally published on Medium.