everything I've built

Workout Coach

A workout app that plans and adapts your training like a real coach would. Personalised strength programmes, warm-ups built around your movement, and a conversational layer so you can ask questions mid-session. In development.

The question

Strength training shouldn't need a PhD in exercise science

Yet most people are stuck choosing between cookie-cutter programmes that ignore them completely, or paying hundreds for a coach who still uses spreadsheets and guesswork. There had to be a better option.

The itch

I've lifted for years and I've tried the apps

The simple ones are just a timer and a log. The complicated ones want video-analysis equipment I don't have. I've worked with coaches too. Good ones are expensive and hard to find. The gap between a generic app and an elite coach is enormous. I kept asking myself: what if AI could bridge it?

What happened

Building from first principles

The starting point isn't "how do we add AI to a fitness app". It's "what would an ideal coach actually do?"

That means real periodisation, not templates. Warm-ups tailored to your specific movement restrictions, not a generic five minutes on the bike. Weights that adjust based on what actually happened last session, not arbitrary percentages. And a conversational layer so you can say "my shoulder feels off today" mid-workout and get a useful answer. That's what I'm building.

AI-generated personalised workout programme showing an Upper/Lower split with exercises, sets and reps Conversational coaching interface with quick-action chips to swap exercises or adjust focus

Early previews: an Upper/Lower split with exercises, sets and reps (left), and the conversational coaching interface with quick-action chips (right).

Join the beta waitlist

This is for people who train consistently but can't justify a personal coach. If that's you, drop your email. I'll reach out when beta opens.

Join the waitlist
A note on how it's being built

Built in Swift with a Go backend. The training engine handles periodisation natively rather than wrapping a generic LLM prompt around a fixed template. The conversational layer uses Claude to interpret in-session queries and translate them into training adjustments. Early alpha is running on my own training data.

  • Periodisation: block structures that adapt based on actual performance, not a fixed week-by-week schedule
  • Mobility: warm-up exercises chosen from your movement history, not a generic routine
  • Load progression: weights recalculate after each session based on what you actually lifted
  • Mid-session chat: ask questions in plain language and get a direct answer, not a generic tip