Products designed for how humans think, not how we wish they did.

I follow a structured AI workflow — a variation of Matt Pocock's 7-phase system — that alternates between human-led decisions and agent-led execution, keeping every session in the smart zone.

1. Idea — I define the problem and desired outcome, then use AI to explore possibilities and refine the concept so we both have a shared design concept.
2. Research — I explore questions and hypotheses using AI to gather insights and validate assumptions.
3. Prototype — I use Gemini with canvas to create very quick prototypes; I also use CLI AI to create higher quality ones for finding issues before implementation.
4. Spec — Using a skill, I convert the idea, research, and prototype into a detailed specification.
5. Kanban Board — I use a skill to translate the spec into actionable tasks.
6. Execution — AI implements each task from the Kanban Board, with a fresh context window for each one.
7. QA — I review the implementation, ensuring tests pass and the user experience is satisfactory.

I start with rough shape and layout using a canvas tool like tldraw or Excalidraw. Once the structure feels right I convert to Penpot or Figma for proper design work, then use AI to generate a clickable prototype.

I prioritise simplicity, usability, and clarity; interfaces should feel obvious in use and honest about what they ask of people.