Skip to content
2026 →ConceptResearch, prototyping

Future AI Products

What I'm exploring next

The problem

Where it started

Most AI products are demos of capability, not solutions to problems. The gap between what models can do and what ordinary organisations actually use is enormous — and it's a design gap, not a technology gap.

The solution

What got built

An ongoing prototype line exploring AI that disappears into the workflow: assistants that draft the boring half of knowledge work, tools that translate between expert language and plain language, and education products that adapt to the learner instead of the average. The prototypes that survive contact with real users graduate into products.

The impact

What changed

This is the laboratory feeding everything else on this site. Prototypes are shared in the AI Lab as they become presentable.

Technology
Claude APINext.jsPrototypingUser research
Lessons learned
  1. 1

    Prototype cheaply enough that abandoning an idea costs nothing but pride.

  2. 2

    The user's first five minutes decide everything; design them before the architecture.

  3. 3

    AI products earn trust by showing their reasoning, not by being right slightly more often.