From "AI Replacement" to Hired in 48 Hours

On May 5th, 2026, I lost my job. My boss had become convinced that AI could do everything I did. Every time a task took longer than he liked, or a bug appeared, he’d hit me with the same question:

"If AI can do your job, where do you position yourself?

But here is the difference between a real engineer and a manager who thinks AI is a "magic button."

The Ghost of Technical Debt

The project that ultimately led to my exit was one I started back in my junior days. After my senior left the company, the codebase sat abandoned for a month. Eventually, I was assigned to assist an outsourced team in reviving the frontend.

Diving back into that repo was a nightmare. It was a rigid cocktail of old LLM-generated snippets and zero linting guidelines. Because I was the only one who could actually parse the logic, I had to refactor it module by module just to make it functional enough to start my actual tasks.

While I was doing the "invisible work" of fixing the foundation, my boss was focused only on the clock. He didn’t see the technical debt; he just saw a slow process. He’d demand to know why the design wasn't perfectly mirrored yet, not realizing I was busy preventing the whole thing from collapsing.

The "AI Ignorance" Myth

The irony is that I wasn’t "AI ignorant." I use AI constantly—so much so that I’m frequently rate-limited. But I knew the tool’s limits, whereas my boss thought the tool was the engineer. When the cycle of unrealistic expectations and "AI can replace you" comments repeated for the hundredth time, I realized that if he thought his AI could do it better, it was time for me to go.

Two Days of Boredom, One Working App

I spent my first 48 hours of unemployment being bored. To stay sharp, I decided to see how fast I could ship a production-ready idea following best practices. I built Space Financial Tracker.

In just two days, I had a fully functional app with secure, subscription-based protection. I used a modern, high-performance stack that requires an actual engineer to orchestrate:

• SvelteKit & TailwindCSS: For a fast, responsive frontend.

• Supabase with Row Level Security (RLS): To ensure data is locked down and secure.

• Shadcn: For clean, functional UI components.

• Payment Gateway: A fully integrated plan and payment system.

Was the UI a bit "AI slop"? Maybe. But the architecture was rock solid, secure, and—most importantly—it worked.

The Ultimate Plot Twist

I was planning to spend day three polishing the UI, but I never got the chance. Within 48 hours of being told I was replaceable, I was scouted and hired by a client in the US.

It turns out that while some bosses are busy looking for reasons to replace their engineers with bots, the rest of the world is looking for the engineers who actually know how to make those bots build something that lasts.