Most automated insulin delivery (AID) systems depend on constant user inputs and heavy manual settings. Rheos is an experimental AID system aiming to change that by using real-time biometric data like heart rate, stress, sleep, and machine learning to automatically adjust insulin dosing.

Instead of traditional carb counting and preset profiles, Rheos allows meals to be logged by photo, voice, or a simple prompt, with insulin needs estimated and suggested to the user. Rheos isn’t alone in rethinking meal logging, as other open source AI-powered photo and voice tools aim to make carb estimation less manual. A built-in large language model also lets users ask practical questions, like how to prepare for a workout or handle a stressful day, and apply those changes with a single tap.

Alongside experimental systems like Rheos, an ecosystem of open-source AID technologies, including Loop, Trio, and AndroidAPS, continues to push automation and customization even further. Rheos fits into this broader movement by learning directly from the body and updating insulin sensitivity throughout the day, with the goal of reducing the daily cognitive burden of diabetes management.

In this episode, we sit down with Jonathan Fitch, founder of GlucoSense, to walk through Rheos, share a live demo, and discuss how his personal experience led him to rethink diabetes automation. Watch the video below or listen above.

Disclaimer: Diabetech content is not medical advice—it’s for educational purposes only. Always consult with a physician before making changes to your healthcare.