I built an AI strength coach because I wanted my training backed by real studies (github.com)

A new open-source AI Strength & Conditioning Coach, dubbed PerformanceAgent, has been unveiled, designed to enhance training by leveraging scientific research. Accessible via an AI agent command-line interface (CLI), such as Claude Code or Gemini CLI, this tool operates without the need for any backend hosting or API keys. Its unique feature is a feasibility engine that evaluates users' training goals, providing transparent, data-driven insights and counter-proposals should a goal be deemed unrealistic. For instance, if a user aims to cut their 10K run time from 55 minutes to 35 minutes in 12 weeks, the coach efficiently calculates a mere 0.2% feasibility for such a drastic improvement. What sets PerformanceAgent apart in the AI/ML landscape is its commitment to evidence-based training, utilizing a deterministic Python engine that ensures the integrity of its calculations and recommendations. By restricting citations to a locally stored, verified evidence corpus, it eliminates the common pitfalls associated with AI-generated content, such as hallucinations or misleading endorsements. The tool’s design allows athletes to log workouts, monitor progression, and receive personalized training programs, thereby establishing a long-term coaching framework that adapts to individual progress. This revolutionary approach not only democratizes access to professional-grade coaching but also underscores the importance of scientific validation in fitness training, which could reshape the way athletes and enthusiasts engage with personalized fitness regimes.