Guides
Why every recommendation has to show its work
A coach that cannot explain a prescription is asking for blind trust. Koyaro is built so the science decides the numbers and the AI only narrates what the engine already computed.
The problem
Black-box coaching is easy to ship
Large language models are fluent. Fluency is not physiology. When an AI invents an FTP, a readiness score, or a fueling target, the sentence still sounds confident — and the athlete cannot check the work.
Serious riders already live with meters, HR, and load charts. They do not need a warmer summary of yesterday. They need a system that will show the math behind today's call.
How Koyaro splits the job
Science decides. AI explains.
A pure sports-science engine owns fitness, fatigue, form, zones, fueling bands, readiness, and load projections. Those numbers are unit-tested against literature-minded values — not improvised in a prompt.
The conversational layer helps you plan, fuel, recover, and race. It is constrained: it can cite the engine and help you act on it. It is not allowed to invent metrics. When you ask “why?”, the answer points at grounded signals — readiness, TSB, CTL, compliance, time in zones — not a vibe check.
That split is the product bet (constrained AI over a deterministic engine). The public methods note lists what the engine actually ships today.
What it feels like
Interrogate the daily call
Transparency is a feature surface, not a tagline. Daily calls, ride analysis, and race tactics expose the metrics they used. You can reshape a session when life changes — and still see how the projection moved.
If a recommendation cannot show its work, it does not belong in the product. That is the standard we hold ourselves to while Koyaro is in closed beta.
Trust is the product.
Join the waitlist and interrogate the first daily call when you’re invited — that is the demo.