About TrainingFlow
Race-day strategy, computed before it is narrated.
TrainingFlow helps marathon and trail runners turn fitness, course profile, and race-day conditions into a practical pacing and fueling plan. The product is intentionally narrow: it does not prescribe training, diagnose injuries, or ask AI to invent performance numbers.
How the engine works
The pacing engine starts from Daniels VDOT or a recent race result, then applies transparent deterministic adjustments for elevation, mean course altitude, heat, humidity, and wind. Nutrition targets come from published carbohydrate, fluid, and sodium intake windows. AI is used only after those numbers exist, and every numeric value in a generated narrative is checked against the source plan before it is shown.
What is live today
The current MVP includes static race pages, profile-based strategy generation, Stripe Race Pass and Annual access, AI narrative validation, and race-week email plumbing. Live Strava import is reserved in the architecture but is not active yet. Open-Meteo race-week forecasts are used when an exact race date is inside the forecast window; otherwise generated strategies use the race climate baseline.
Safety boundaries
TrainingFlow treats athlete profile data as sensitive health-adjacent data and stores it in Supabase's EU region. The app is a race-day planning tool, not a medical or coaching replacement. See the privacy policy for data handling and the pacing guide for method details.
Who builds this
Kristian Hoffmann
Founder — Hannover, Germany
Kristian builds TrainingFlow and is responsible for the pacing engine, the race catalog, and the guides published on the blog. The formulas behind the engine are documented on the accuracy page, including where they come from and how they are tested.