Thorax

Context

After years of lifting, I was still arriving at the gym without a clear structure. Existing apps could log sets, but none gave me the coach I wanted: one that understood my training history, respected recovery, and could adapt a session when equipment was unavailable or something did not feel right. So I decided to build it from the beginning.

How I approach it

I designed Thorax as both a system of record and an interpretation layer. It brings routines, live workouts, progression, recovery, body composition, sports, and community into one product. Conventional metrics remain deterministic; Thorax AI uses specialised tools to inspect the user’s real data, explain patterns, and propose useful changes. Any action that changes data is presented for explicit confirmation first.

• Routines, live workouts, progression, recovery, body composition, and sports in one platform
• ThoraxGPT: a data-grounded coach with five personalities and specialised analytical tools
• Deterministic metrics and controlled AI actions with explicit user confirmation
• Health Connect and Apple Health data feeding readiness and performance context
• A multilingual TypeScript product shipped across web, iOS, and Android

Why it matters

Thorax brings together the areas I care about most: product judgement, analytical thinking, data modelling, AI systems, interface design, and end-to-end execution. It shows how a personal frustration can become a coherent product that people can use in the gym, even with poor connectivity, rather than another isolated prototype.