About

The coaching years

I coached rowing for eighteen years. Crews I worked with won seventeen medals at international regattas, four of them gold, and went two seasons unbeaten in Europe.

For the first seven of those years I had a problem I couldn't name. In training my rowers improved steadily; at races — a mile from the coach, in pain, unable to hear anything — they imploded. Everything built in controlled conditions evaporated under pressure. I was giving technically accurate instruction, lots of it, and it wasn't transferring.

Two conversations changed how I coach. The first was with Rusty Williams, a legend of the sport, after a session that was going badly. He asked me to explain how to row. Somewhere around my thirtieth technical point he cut me off: "You put the oar in the water, push hard, and take the oar out of the water." Not a correction — a compression. Three things the athlete's body could work out how to do, instead of thirty it had to consciously manage. Seven months later my crew won gold at two international regattas.

The second came from Ashley Neeves, one of the UK's leading tennis coaches, who taught forehand mechanics with the phrase "show the logo, hide the logo". I reversed it for rowing — hide the logo on your shirt through the drive, then show it — and a movement pattern that had taken me nine to twelve months to teach, with about a 60% success rate, started taking 90 to 180 minutes with close to a 100% success rate. Six words did what hundreds of instructions couldn't.

What that taught me

Those two stories are the same lesson. Expert knowledge is tacit: the people who have it genuinely cannot say what it is, so they teach a vague description of it instead — and when the description doesn't transfer, everyone blames the student. Getting skill to move between people takes deliberate work at both ends: extracting the pattern from the expert, then encoding it into cues a learner's perception can actually use. I've written up the ideas that came out of this — start with the double translation problem.

Building things

I founded Afterlife Carbon, a venture that intercepts end-of-life carbon-fibre sports equipment and turns it into premium consumer products with verified provenance. It took me into rooms with British Rowing, SailGP and Cervélo, and taught me a different kind of craft: positioning, provenance and brand. There's more on the building page.

More recently I've been running experiments at the junction of AI and expertise: a structured knowledge system that interviews me the way a cognitive task analyst interviews an expert, and two complete companies designed autonomously by an AI agent working from that system. The point of both is the same question the coaching years posed — how much of what a person knows can actually be captured, and what happens to the part that can't.

Now

I write and think about how expertise forms, why it resists transfer, and what generative AI does to both — the flagship piece is here. If any of this is your territory, I'd like to hear from you.