It learns your habits
It watches how your recovery, measured by your morning HRV, reacts to the load of the previous days. Every cyclist responds differently: the model adapts to you.
Early access · Cycling
Load a route, choose the power you want to ride it at and see in advance how your body will respond: heart rate, anaerobic reserve W' and the points where you risk blowing up.
The What-If simulator
A GPX file of the route, or one of your saved routes.
Watts for every section: climbs, drags, descents. The curves update while you edit them.
Expected heart rate, W' depletion and the critical points where your reserve runs out, with advice on how to handle them.
Training that fits you
Digital Twin is connected to a machine learning system that learns how you respond, not an average athlete. Before every session it tells you whether, and how much, to change it.
It watches how your recovery, measured by your morning HRV, reacts to the load of the previous days. Every cyclist responds differently: the model adapts to you.
Based on your current form and the workouts you have already completed, it suggests whether to keep the planned session, ease it off or step it up, and by how much.
With "Chiedi all'IA" (Ask the AI) you get the suggestion right on today's or this week's session, with the reason. If there is not enough data it tells you, instead of making something up.
During early access, the AI suggestions are labelled as demonstrations: the final decision stays with you or your coach.
For coaches
Digital Twin has a Coach mode, in the same app: connect your athletes and work on their data as if they were in front of you.
The athlete shares a code with you from their profile and decides what you can see: they can revoke access at any time.
Under the hood
The calculation is based on critical power (CP) and W' and on a model of the heart rate response. Every formula is documented, has declared units and is checked by tests against hand-calculated values.
With an account, the app imports your activities from intervals.icu and calibrates the digital twin on your recorded power, heart rate and HRV.
The engine runs inside the app, so the curves update instantly as you change your targets. The server only stores the data.
Machine learning runs in a separate service and never touches the formulas of the physiological engine. Until there is enough data to validate it, the app presents its suggestions as demonstration estimates.
The project
Digital Twin combines physiological models and training data to predict how you respond to effort.
It is a demo, not a medical product: the forecasts are for exploring scenarios, not for replacing a coach or a clinical assessment.
The app includes an anonymous demo dataset. If there is no connection, the simulator still works.
The site uses Google Analytics, with cookies, only if you accept. Your training data stays in your account and is not shared.
Get the app
The app can already be downloaded as a preview: on Android from Google Play, on iPhone through TestFlight, Apple's app for betas (it's free on the App Store; the link guides you).
You need an invitation to get in: if you don't have one, request early access from the app itself or with the form below.
Sign up
We are opening the app to a small group of cyclists, coaches and researchers. Request early access to try it now and help us improve it, or sign up for the launch notice and we will write to you when it is available to everyone. The app is currently in Italian; an English version is planned.
We will send you an email to confirm your address. We will only use your data for this, and you can unsubscribe at any time.
Can't see the form? That happens in reading mode: tap the document icon to go back to the full page, or write to info@digital-twins.app to request access.