Adaptive planning guide
How adaptive hybrid training plans work
Keep athlete data, the full plan and completed history in one system; let AI assist with bounded proposals while deterministic rules decide what is safe to apply.
How is this different from asking a general AI chat for a plan?
A structured product keeps validated athlete inputs, the full progression, completed sessions and decision history together instead of relying on a long prompt thread.
General AI chat is useful for exploring ideas. Ongoing training becomes harder when every update requires re-pasting context, reconciling old answers and manually tracking which version of the plan is current.
Octalap stores bounded athlete context, materialises workouts and records accepted changes. The database and deterministic engine remain the source of truth; raw prompts are not the training record.
What information improves an adaptive plan?
Race goal, recent benchmarks, normal weekly availability, equipment and completed training provide the core context; body metrics and richer profile details remain optional.
Measured benchmarks improve pace and load suggestions, but a beginner can start with estimates and add tests later. Optional context should explain exactly what it improves rather than blocking first value.
Weekly adaptation needs only coarse recovery and schedule inputs plus the work actually completed. It does not need a wearable score or a diary of private medical information.
What does AI decide?
AI may help structure or explain a proposal. It cannot bypass the plan schema, rewrite completed history or override load, schedule, equipment, deload and taper rules.
Every generated surface identifies whether it was calculated by rules or AI-assisted and validated by rules. Invalid, refused, timed-out or unavailable AI output falls back to the deterministic result.
The athlete sees a proposed change before accepting it. Ask responses that imply a schedule change become a confirm-before-change proposal rather than a silent edit.
How does the plan adapt after a difficult week?
It changes only uncompleted future work and keeps the next week inside conservative load and availability boundaries.
Normal next-week load is capped against recent completed load. Poor sleep or elevated fatigue and soreness lower the target; programmed deload and taper weeks retain their reductions.
The decision log records the before/after sessions, minutes and reason codes. That makes an accepted adaptation inspectable and keeps a later model response from becoming an undocumented source of truth.
See where to start
Use the two-minute readiness baseline, then explore the sample product before deciding whether to create an account.