This is a research experiment exploring how a Large Language Model (LLM) can create personalized fitness programs under professional coaching supervision.
The project focuses on a specific case study: an amateur, master athlete practicing CrossFit® style training, for personal wellbeing, not competition. The AI has been trained using open-source material from the CrossFit® Journal and Training Guides to understand programming principles and methodology specific to CrossFit® in addition to general fitness exercise and theory articles from various recognised sources such as PubMed, NCSA, Cathalist Athletic and others, .
The experiment aims to investigate how effectively an AI system can:
- Interpret an athlete’s specific needs and limitations
- Create appropriate training programs based on functional fitness principles
- Adapt programming to a master athlete’s requirements
- Generate workouts that balance intensity with sustainability
Every program generated by the AI undergoes a coach review to validate its appropriateness and effectiveness. This human oversight ensures training quality while allowing us to study the current capabilities and limitations of AI in fitness programming.
Programming is developed on a week-by-week basis, allowing for continuous adaptation based on two key feedback sources: the athlete’s daily subjective feedback and objective data collected from wearable technology.
In particular utilize data from a Garmin® top range smartwatch and heart rate monitor, along with their associated software, to track vital metrics such as training load, recovery status, and physiological responses to workouts. This dual feedback system enables the AI to adjust and optimize the programming based on both perceived and measured responses to training.
This is not a commercial product but rather an academic investigation into the potential of AI as a tool for fitness program design when properly trained and supervised.
We’ve decided to share this experiment openly to engage with the fitness community and gather valuable feedback. If you have questions about any aspect of this project – from the AI training process to specific programming choices – we’d love to hear from you. You can submit your queries through our contact form, and we’ll be happy to discuss the experiment in detail.