I designed buildings for six years before I designed interfaces.
The method didn't change. The medium did.
Before pixels, I spent six years on parametric architecture. Parametric design isn't about drawing lines; it's about setting mathematical rules, modelling constraints, and letting systems generate the geometry. That exact mindset is why I excel in the 2026 AI market. LLMs and autonomous agents are incredibly powerful, but they lack human context, behavioural empathy, and spatial reasoning. They need a blueprint. My architectural training lets me treat AI not as a magic wand, but as a computational material. I don't just prompt algorithms to I architect system constraints and define behavioural thresholds.
Today I design behavioural UX: insurance journeys where people are in actual distress, or AI assistants that must earn trust before being useful. Because I understand structural logic, I can translate messy human psychology into strict parameters for AI models. I use Gemini and Claude to stress-test system architecture, Claude Design and ComfyUI for high fidelity visual pipelines, and orchestrate agents via Google Antigravity, Replit, and Lovable to spin up live React applications. AI builds the house; I design the blueprint.
Outside client work, I teach the anti brief method to invert what a client asks for and surface the problem it's hiding to and run Zaltt Labs, where I publish unfinished thinking. No case studies. No polish. The actual experiments, honestly documented.
