✨ Vibe coding: building web and mobile apps with AI (my hobby outside work)
June 1, 2026 - by Tom de Bruijn 3 min read
Outside my work with clients, I spend my free time on something often called vibe coding.
What is vibe coding
Vibe coding is building software by describing what you want in natural language to an AI coding agent. Instead of writing everything yourself, you work in iterations with an AI model. You define the idea, direction and quality bar. The AI generates and adapts the code.
With tools like Claude Code, Cursor and GitHub Copilot, you can go from idea to working web or mobile app in a short time. This website itself has been partially built and improved this way.
What vibe coding really is in practice
On the surface it looks simple: you describe something and code appears. In reality it feels more like working with a fast engineer that only performs well if you keep the direction sharp.
Without clear structure, the model will make its own decisions, which often leads to unnecessary complexity later. Vibe coding only works well when you actively control architecture and scope.
Where it often goes wrong
The biggest issue is not bad code, but growing complexity without design. A project starts small and clean, then gradually becomes something where decisions are implicit. It still works, but becomes hard to understand and change.
Token usage is a real constraint
AI work does not scale freely. Everything in your context affects the next step. Long conversations and broad context make models less consistent.
That is why I use tooling that provides insight into token usage and context optimization, such as RTK and Token Optimizer. Not just to reduce cost, but to keep context clean and focused.
From hobby to production
Vibe coding works well for prototypes and experiments. Once something moves toward production, different concerns become important: reliability, security, maintainability, dependencies and ownership. This is a different domain from the vibe coding workflow itself.
In my work I help organizations bridge that gap: from fast AI prototypes to production-grade systems.
👉 More about this: see my AI Implementations offering.
At that point it is no longer about "building something with AI", but about architecture, governance and integration into existing systems.
What I am building
During the World Cup I built a tool that predicts match outcomes, aiming to outperform bookmakers over time. The tournament has ended, but the experiment took flight: it hatched Oddsprey, the project I am actively building on right now.
I am also exploring AI-assisted game development, mainly small mobile games built from idea to playable prototype very quickly.
📬 Want to exchange ideas or experiences? Find me on LinkedIn or get in touch via the options below.