What I'm on,
right now.

Most of my attention is on making useful things with AI, then finding out whether they’re actually useful. The second part takes longer than the demo.

Building

I’m working on software, prototypes and practical ways to use AI in everyday work. I’m particularly interested in giving agents a clear goal, letting them handle more of the steps, and checking what comes back.

Agent teams are part of that exploration: when separate roles help, when they add overhead, and how to tell the difference before assembling a committee to fix a button.

I’ve also rebuilt this site and added a writing studio. It’s now easier to turn an idea into a draft. Whether that makes me publish more regularly remains an excellent question.

Writing

I’ve just published Give It a Goal, a piece about the different ways I use AI—from asking questions and tidying up sentences to building things with teams of agents.

The thread running through it is how to hand over more of the work while staying responsible for the result. I’m trying to make that practical, personal, and readable without a glossary.

Thinking about

What becomes more valuable as software gets easier to make.

My attention keeps coming back to judgement: choosing a worthwhile problem, explaining it clearly, and recognising when the result is good enough. Those decisions are easy to overlook when there’s something impressive appearing on screen.

I’m also trying to get better at finishing the thing I started before asking, “Could you also…”