About

I'm consulting for the UK government's AI Studio at GDS, helping bring AI agents into public services. The question I keep turning over is how to make those agents legible — readable to the citizens they serve, and to the institutions they speak for.

In 2014, ahead of its time, I co-founded Normally, an applied AI studio. We worked with Apple, IKEA, Google, Panasonic, the NHS, and the BBC, turning data into products, platforms, and services. Because the material was new, we built the practice to match. R&D-first and just-in-time process. We built interdisciplinary hybrid teams, a collective growth culture, and our own tools for designing with data and AI — all on a four-day working week, instituted before the idea had a name.

In 2010, I bought the entire Companies House dataset for £14,000, ran nine million scanned filings through a pipeline, and put a public-facing interface over the top. I think that work helped push the data toward becoming free and digital.

In 2001 I co-founded Livework, the world's first service design company, where I helped establish the discipline internationally. In the same year, I sold my own personal data on eBay as a design provocation about ownership and consent. Around that time I also contributed to the founding of the Interaction Design Institute Ivrea.

Earlier still: in 1995, in my final year at Glasgow School of Art, I co-designed the world's first online degree show.

I've taught throughout my career — at Imperial Business School, CIID, Glasgow School of Art, and Ivrea.

Taken whole, it's one piece of work. The through-line is legibility: making invisible systems readable, contestable, navigable. Service systems at Livework. Data at Normally. Now agents. The thread doesn't change — only the material does. Make the thing visible, and the people inside it can act.

Throughout, the quixotic quest has been the same: build for the attentive, not the attention economy.

Nescient is the foundation. The word means "not knowing," but in the older, more careful sense — principled not-knowing, Socratic rather than ignorant. I think the most useful posture toward AI right now is exactly that: a curiosity that doesn't pretend to know the answer before the question is finished. These systems are unfamiliar materials. They reward attention, patience, and the willingness to be wrong out loud. They punish the opposite.

This site is where I publish what I'm working on and thinking about. Some of it's finished. Some of it isn't. I write in the open because I'd rather think with people than at them.