I Built a Production App in 3 Weeks. Alone.

A few weeks ago I shipped NeonLoops — a visual platform for building AI agent workflows. Drag-and-drop canvas, 12+ node types, structured outputs, code export, the whole thing. Three weeks, start to finish, solo.

That would’ve taken a small team months not long ago. It took me three weeks and Claude Code.

What actually changed

I’ve been a software engineer for nearly a decade. Worked at a bank, a major telecom company in Turkey, now at a logistics startup in Vienna. I’ve seen enterprise codebases, legacy systems, the whole spectrum. I know what production code looks like.

That experience didn’t go away — it just got amplified. I don’t write most of the code anymore. Instead, I:

Architect — decide the structure, pick the stack, design the data model
Describe — tell the AI what I need, clearly and precisely
Validate — read every line, catch the mistakes, push back when it’s wrong
Ship — deploy, test, iterate

The bottleneck used to be “can I write this fast enough?” Now it’s “do I know what good looks like?” And after years of building software, I do.

The NeonLoops story

I had the idea on a Sunday. By Wednesday I had a working canvas with React Flow. By the end of week two, I had auth, a database, structured outputs, and a preview panel. Week three was polish, testing, and deployment.

Week 1: Canvas, node system, basic flow execution
Week 2: Auth, database, structured outputs, preview
Week 3: Code export, SDK generation, polish, deploy

Every single day I was shipping features that worked. Not prototypes — actual production features with error handling, edge cases covered, tests written. The AI wrote most of the code. I made sure it was good.

The skill that matters now

Here’s what I’ve learned: the developers who thrive with AI aren’t the ones who prompt the best. They’re the ones who’ve built enough things to recognize when something is off.

You still need to read every line. You still need to catch the subtle bugs. You still need taste.

When Claude generates a React component, I can tell in seconds if the state management is right, if the error boundaries make sense, if the accessibility is there. That comes from years of writing bad code, then better code, then learning what “good” actually means.

My head won’t shut up

The real problem now isn’t ability — it’s time. My head is constantly buzzing with ideas. I’m rebuilding MysticLab with my wife Mihrimah (she built the original dream interpretation app, now we’re tackling the next version together). I’m working on Stick Story. I have a fitness app, a security tool, language learning systems — all in various states of existence.

I don’t have enough hours in the day. But I keep building anyway, because I genuinely can’t help it. It’s not about money (none of these make money yet). It’s a drive — like a reflex. An idea shows up, and I have to chase it.

What I am now

The title “software engineer” doesn’t fully capture it anymore. I’m part architect, part editor, part product person. I design systems, validate AI output, and ship products that look and feel like they were built by a team.

I moved from Turkey to Vienna three years ago. I lift weights in the morning, stretch (a surprisingly life-changing recent habit), take long walks with Mihrimah, and spend my evenings turning ideas into real things.

I don’t know what to call this new role. Builder, maybe. That feels closer.

This site exists because I needed a place to point to and say: “here’s what I make.” If you’re reading this, now you know.