Week-by-week training plan from a deterministic periodisation model. No language model touches the numbers.
Rendered directly from the constants the engine uses, so this table cannot drift from the code.
The numbers above are settled. This is where a language model earns its place: turning a validated plan into something an athlete will actually follow. The button copies a complete prompt with the plan embedded. Paste it into Claude or ChatGPT and you get the rationale back. No key, no server, and the model is handed numbers it cannot change.
These are widely used coaching conventions, not settled science, and their evidence base is uneven. The taper rules are the best supported of the set. The 10% weekly progression cap is closer to folk wisdom that has survived because it errs on the safe side than a demonstrated threshold. Sessions are prescribed by effort rather than pace, because no pace or heart-rate data is collected here and any pace printed would be invented. This is not medical advice.