According to Andrej Karpathy on X, he is experimenting with a personalized, regimented software workflow to lower his resting heart rate from 50 to 45 bpm, illustrating a near-term shift toward highly bespoke AI software that adapts to individual goals and biometrics. As reported by Karpathy’s post, the experiment highlights AI-driven coaching loops that integrate wearable data, micro-targeted protocols, and continuous feedback for outcome optimization. According to the post, the practical business implications include verticalized AI agents for fitness and cardiometabolic health, subscription coaching models linked to biomarker targets, and integrations with wearables and EHRs for measurable ROI. As reported by Karpathy, this approach underscores demand for model architectures that support user-specific objective functions, fine-tuned habit formation nudges, and automated experimentation frameworks, creating opportunities for developers to build closed-loop health agents with compliance tracking and outcome guarantees.