AI as a working partner.

AI has genuinely changed what one person can build, analyse, and ship. Things that used to need a team — data pipelines, localised content at scale, functional prototypes, server infrastructure — are now within reach of a single person. This portfolio is a decent example of that.

The catch is that the tool getting more powerful doesn't mean it's getting wiser. I think about it through two analogies. A calculator given a pen — because I'm asking a computational model to do creative work, which means I need to stay close to what it's actually producing. And an autopilot given the cockpit — because no matter how capable the system, the final call has to be human. Both feel true at the same time.

The industry has a term for what happens when people stop paying attention — comprehension debt. It's real, and I've seen it. I haven't written production code in years, but when I build with AI I make sure I understand what's happening well enough to debug it when something goes sideways. And something always goes sideways eventually.

Here's what I've actually built.