who I am, what I build, why I build it
I'm Dax. I design systems — I don't code them. I define the architecture, the constraints, the philosophy. Then I delegate execution to my models. The tagline is "architecte non-codeur" and it's not a euphemism. I've never written a production codebase. I've built a complete local AI agent stack, a distributed multi-agent system, a custom quantization pipeline, and this website — all by describing what I want to the right model in the right context.
Based in Zwevegem, Belgium. The infrastructure runs from a 14-liter ITX case in my living room.
My primary agent is pi — a lean, vision-capable agent distilled from Claude Opus reasoning traces, running on a 27B dense model. It has four core tools (read, write, edit, bash), native vision, persistent memory, and the ability to escalate to frontier models when it hits its reasoning limits. No bloat, no permission prompts, just focused execution.
A sovereign AI stack. Local inference, local memory, local decision-making. The goal is simple: own the reasoning system you depend on. Not rent it, not subscribe to it, not hope that the provider doesn't change the terms next quarter.
I apply Popperian falsifiability as a method — to systems, to architectures, to models, and to my own thinking. If a claim can't be tested, it's not a claim. If a model can't say "I don't know," it's not honest. If a system can't fail gracefully, it's not resilient.
The local AI stack isn't about privacy as ideology. It's about controlling the variables that matter. Local inference means the model's behavior is reproducible. Local memory means the identity scaffold is mine to shape. Local search means my queries don't train someone else's model. The rest — Gmail, YouTube, cloud backups — stays where it is. Pragmatism beats purism.