Writing
Blog
Thoughts on AI, sports science, performance modelling, and the intersection of research and engineering.
The descent from the Poggio is where Milano-Sanremo is won or lost. A look at what makes those 3.5 km so decisive — and what the science of bike handling and optimal trajectories can tell us about it.
Spragg et al. 2023 aligns perfectly with the Athletica Workout Reserve concept — a real-time mechanical model that monitors how an athlete's maximal mean power shifts as fatigue accumulates.
What does a hunting eagle have in common with a sunflower? Both follow Fibonacci spiral patterns — and experienced cyclists may trace clothoids in their descending lines.
A data-driven analysis of the Verona ITT course comparing the 2019 and 2022 editions, exploring how racing lines and descending strategies affect time trial performance.
Most AI algorithms for CPET data never reach clinical practice. What is the AI chasm, and how do interpretability, transparency, and cognitive load shape whether a model survives contact with the real world?
Bike handling is a fascinating and under-studied topic in road cycling. I define it as the ability to consciously explore large portions of the gg diagram — and it can be worth over a minute in a 5-km technical section.
How a crowd-sourced dataset of expert annotations combined with deep learning can automate and standardise CPET interpretation at scale — while tackling the AI chasm from a different angle.
How AI technologies might assist us to monitor and optimise the chronic physiological adaptations that occur with endurance training, and how AI models might enable more systematic, data-driven training design.
We didn't need electric guitars to make better music — and we don't need AI to become better coaches. But we needed electric guitars to create new sounds. AI in sport science will be similar.