Building with AI
Independent projects in AI-assisted development, validated with experience.
I use AI as an active part of my development process: exploring solutions, building prototypes and accelerating implementation. The projects below demonstrate how I apply AI-assisted development to real problems, from handling complex datasets to implementing algorithms where correctness matters. My focus is not just on what AI can generate, but on turning that output into reliable, working software.
Selected work

AI-first development · Technical product owner · Logical thinking
Logic puzzles and teaching how to solve them
Puuzel.nl offers logic puzzles, with solvers, puzzle generation, hints and step-by-step tutorials that teach the reasoning. That is behaviour that is easy to make plausible and hard to make correct: every generated puzzle must have exactly one solution, and every hint must follow from the board.
I deliberately started with few strict requirements, using the project to explore how far AI could take an idea with minimal upfront specification — and how much guidance it needed along the way. AI designed and implemented the algorithms for all six puzzle types within a generic, reusable system, and produced extensive unit and content tests. It also became part of the development workflow, including managing pull requests and deployment.
My role was primarily that of technical product owner: evaluating what AI produced, deciding where to steer it, and judging whether the result worked as a product. Were the hints logically sound and intuitive? Did the tutorials explain the reasoning clearly? Did the difficulty settings reflect the actual challenge? The project was a practical exploration of AI-assisted development, balancing the freedom to let AI find its own solutions with the technical and product judgement needed to make the result reliable and useful.

Collaborative AI development · Heterogeneous open data
An interactive map of the trees of the Netherlands
Bomenatlas.nl brings together municipal tree inventories, mostly published as open data, with information about tree species. The challenge was integrating data from different sources, with varying formats, levels of detail and data quality, into a coherent and reliable whole.
Unlike Puuzel.nl, where I deliberately gave AI room to find its own direction, this project involved a more collaborative, side-by-side development process. We worked through the practical challenges of importing, normalizing and validating heterogeneous datasets, making the resulting information consistent and useful despite the limitations of the source data.

Pet project · UX focus
A block puzzle game for my daughter, without the ads
Block Blowup started at home. My daughter likes to play a block puzzle game on her phone, but the ads between rounds kept spoiling the fun. So I had AI build her an ad-free version of her own, installable as PWA.
For me, this little project captures two aspects I regularly experience when developing with AI. On the one hand, writing the game myself would have been genuinely satisfying; I enjoy that kind of coding. On the other hand, letting AI take care of the implementation freed me to spend my time on how the game looks and feels.