Goals for 2026

What would happen if I treated ChatGPT seriously as a coach?

Not as a novelty, but as something that would enable me to build a structured, progressive plan and help me with things like pacing and nutrition. I assumed the models must have been trained on huge amounts of relevant content, this is an area where a ‘probable’ answer is what you want.

I began using ChatGPT as an endurance coach. I fed it my data, asked it to interview me and got it to build me a plan.

The experience was often clunky and laborious. I had to resort to converting fit files to CSVs, I dropped in screenshots from Strava and manually added my metrics but the advice seemed solid and aligned with everything I’d learned about endurance training. It quickly became obvious that this wasn’t just an interesting experiment. I was getting fitter and learning more about what good training looks like.

Fast forward a few months and something unexpected happened.

Ed Crossman came to me with an idea: building an AI-powered coaching app for endurance athletes. When we compared notes, it was clear we were circling the same insight from different directions. Zepho was born.

We knew we were onto something when it became clear that I wasn’t alone. We found more and more people that were quietly hacking together prompts, and ad-hoc systems to make ChatGPT behave like a coach. While they were impressed with the results I could see their experience was fragile, manual, and poorly designed. We realised we could craft an experience around this awesome capability by creating a delightful and intuitive bespoke context collector.

For me, Zepho sat immediately at the intersection of three long-standing interests: design, startups, and endurance sport.

As a designer, I was drawn to shaping new patterns powered by this incredible new capability. And as an athlete, to whether this would actually hold up under real conditions.

Fast forward to June and it was time for the nationals. I felt great, I was the fittest I’d been in years. I entered a few events, came 3rd in the scratch race and somehow managed to win Gold in the Kilo.

The general training plan was solid but it also helped in some unexpected and interesting ways. Such as a bespoke recovery and nutrition between two events on the same day (a scratch race in the morning and the kilo in the afternoon). I don’t think that I won that race because of an AI coach, I appreciate that correlation is not causation and that I’m a sample size of 1, but it’s certainly spurred me on to take Zepho to the next level.

I don’t want to blow my own trumpet, it’s a niche sport and it takes a certain mentality to want to spend a weekend at the Newport velodrome. But I’m still pretty chuffed with winning a national jersey.

Which (after quite a long back story) brings me back to now.

This year I’ve got some even bigger plans. Not only will I be returning to the Masters Nationals in June, the Masters World Championships is on my doorstep at Lee Valley in October, so it seems rude not to give that a go.

But first on my list is launching Zepho. Let me know if you’d like to try the beta, if not keep your eyes on the app store in the coming weeks.