Building with AI

For most of my career, an idea had to clear a high bar before I could act on it. It needed to be important enough to justify a team, a roadmap and months of engineering time. That left a long list of small, useful products in my head: things I would use every day, but could never quite justify building.

AI changed the maths. I can now make a real version quickly enough to find out whether an annoyance is actually worth solving. These apps are not concept work. Each one has forced me to learn the parts of product development that used to sit just beyond my reach: authentication, bank connections, health data, OAuth, native platforms, distribution and the unglamorous work of earning trust.

The collection now ranges from personal finance and cooking to language learning, local discovery and a personal assistant. Each product has its own home on the web, where the idea can be understood and the working product can be opened directly.

I am not building for the sake of building. The point of the exercise is to practise judgement. When making becomes cheap, it exposes the part that was always difficult: deciding what deserves to exist, what should be left out, and what needs to be true before a product is genuinely useful.

Each app gives me another set of reps across hundreds of small decisions—where to begin, what to simplify, which edge cases matter, how the interface should explain itself, and when the value is real enough to stop. Once I reach that point, I move to the next problem. The apps are the output; the compounding skill is judgement.