I started coding at 15, a World of Warcraft website that got DDoS'd on day one. By 18, I was building bots and learning how production systems resist external interaction. After university (where I'd been learning to build e-commerce sites and PHP), I took a different path: Java development in a regulated sector in Luxembourg. Three years at Finologee building systems for the banking sector, where every deployment is audited and every mistake has legal consequences. That's where I learned that production-grade isn't a feature. It's a discipline.

In 2020, I co-founded kodehyve as CTO. The mission: build the operating system for real estate developers: a multi-module, multi-stakeholder platform covering project management, CRM, financial management, compliance, and investor lifecycle. The end goal: manage everything from land acquisition to key handover.

I was the sole developer for the first year. I wrote the code, chose the stack, designed the architecture. As the company grew, I built the entire development process: environments, CI/CD, code review workflows, onboarding, training programs.

I scaled the dev team to 10+ engineers, introduced event-driven architecture, built a shared design system, and led a complete platform rebuild when we outgrew the original codebase, because rebuilding was faster than fighting the debt we'd accumulated under pressure.

On the side, I built a fully eIDAS-compliant eSignature platform using Luxembourg's national trust infrastructure (LuxTrust) as the underlying provider. It's used by 100+ real estate agencies today.

5+ years of building at this scale (from sole developer to a team of 20+) taught me that every foundational decision compounds. The ones you get right carry you. The ones you get wrong haunt you.

And then AI arrived. Starting in 2025, I spent entire days testing, not building demos, but trying to find out: can it really replace a developer? What kind of developer? Is it hype, or is it actually as wild as people say? The answer: without guardrails, it creates beautiful chaos. With solid foundations and defined patterns, it becomes the most powerful tool I've ever used.

And that's when it hit me: it's like having employees who can do anything. But you have to explain everything to them. The best employee in the world, except they don't give a damn about your business. And because they love you, you're always right.

That's where ITzWorking started: an AI-native dev agency, where the years I spent shipping production systems become the frame AI doesn't have on its own. I spent months isolating the pillars that turn an LLM from a 2x intern into a 10x senior. I wrote them down, clarified them, stress-tested them on real projects.

From those pillars, two processes emerged: one to integrate AI-assisted development into an existing codebase, another to ship a complete project alone in seven days, work that would have taken five developers six months. Real numbers.

What I learned along the way: every piece of software faces the same underlying problems. You only recognize them once you've lived them. Read more at itzworking.io...

Coming out of kodehyve, I had something rare: a CTO's view of what actually slows a company down, and the execution speed of an AI-native dev agency. Watching whole teams burn out on repetitive work, the realization landed: those teams can be replaced or multiplied by AI. Not by stacking ten SaaS tools. By building their own software.

It hit me hardest while building Nexus, ITzWorking's internal tool. Early on, everything lived in notes and human memory. Then we did the dance everyone does: Notion, Linear, Slack, HubSpot, Google Drive: ten tabs and a constant context tax. Nexus pulled all of it into a single interface: project management, lead gen, people tracking, sprint planning, proposal drafting, client responses. First viable version online in a day. By day five, the gap with our old way was obvious.

story.chapters.itzjulien.paragraphs.2

story.chapters.itzjulien.paragraphs.3

Identifying that leverage isn't a workflow or a methodology you copy-paste. And the AI that comes out of it doesn't replicate on an assembly line either. That's why it's artisanal: like a hand-made object, you'll see similar shapes from one client to another, but never the same tool. Always tailored to the company that carries it.