Training cycles generation
A practical guide to getting the best structured periodization plans from the AI Assistant — what it knows, what drives its decisions, and where your input shapes the plan most.
This feature is designed as a support tool for the coach, not a replacement — we believe that only a plan reviewed and refined by an experienced coach truly meets the quality bar that athletes deserve.
What the AI Assistant produces
Each generation creates a multi-week periodization plan structured as a flat list of interlocking cycles:
- Macrocycles — the large training blocks (typically 8–15 weeks each) that cover the full requested period without gaps
- Mesocycles — the focused sub-phases within each macrocycle (typically 3–6 weeks each), filling their parent block completely without overlaps
Each cycle includes a name, purpose description, and date range. Colors for visual identification in Good Coach App are assigned automatically. The plan is built around your athlete's goals, known competitions, and your coaching methodology.
The generator does not produce daily workouts. It gives you the periodization architecture — the structured framework of phases that shapes all training decisions within each block.
The two-step flow
Training cycles generation is a review-before-commit process. It works in two phases:
Phase 1 — Draft (text plan)
The AI Assistant generates a readable text plan describing the proposed cycles, their rationale, and the overall periodization strategy. This is shown to you for review — nothing is saved to the athlete's calendar yet.
Read it as a coach reading a colleague's proposal: does the structure make sense for this athlete, are the phases in the right order, do the dates align with competitions and goals?
Phase 2 — Feedback or acceptance
After reviewing the draft, you do one of two things:
- Accept it — submit "Looks good" or any brief confirmation. The AI Assistant will finalise the plan as structured cycles and save them to the athlete's calendar.
- Request changes — describe what you want adjusted. The AI Assistant revises the plan while keeping what worked, then saves the final version.
Either path goes through the feedback step. This is by design — it ensures cycles are only created after a coach has reviewed them.
Phase 3 - Coach adjustments
After the plan is saved, you can make any manual adjustments in Good Coach App — renaming cycles, changing colors, adjusting dates, or even restructuring the phases. The AI Assistant's output is a starting point, not a locked-in final product.
What the system knows automatically
Before generating, the system loads context from both the athlete's profile, accumulated athlete memory and coach's accumulated coaching history. You don't need to repeat any of this in your inputs.
| What | How it's used |
|---|---|
| Coach memory | Your coaching methodology, how you structure blocks, session preferences, and periodization patterns built up from past generations — see the Memory section |
| Athlete memory | The athlete's long-term response to training: what phases they handle well, injury patterns, what worked or didn't across previous training cycles — see the Memory section |
| Existing training cycles | Current macrocycles and mesocycles in the athlete's calendar — used to avoid overlap and understand where the new plan fits |
| Upcoming competitions | Competitions you include in the request (see below) directly shape the phase structure and taper placement |
Read more in memory section.
Required inputs
Start date and weeks count
The plan covers exactly the number of weeks you specify, starting from the date you provide. The minimum is 10 weeks — shorter periods don't give the AI Assistant enough room to build a meaningful periodization arc.
Start date should typically be a Monday, as all cycles start on Mondays.
Goals description
This is the most important input for the AI Assistant as the primary brief — it determines what phases are needed, what qualities to develop, and in what order.
What makes a good goals description:
| Less effective | More effective |
|---|---|
| "Improve fitness" | "Improve lactate threshold pace from 5:30/km to 5:00/km over 16 weeks, with a local half marathon in September as a target race" |
| "Run faster" | "Prepare for a 10k PB attempt in autumn. Athlete has a solid aerobic base but lacks speed-endurance. Has done no structured threshold work in the past year." |
| "Stay fit" | "Maintain aerobic base through summer with minimal competition stress. Athlete returning from a 6-week injury break — conservative progression expected." |
Include:
- The athlete's current fitness state and what's lacking
- The target outcome — a specific race, a performance metric, a fitness quality
- Any relevant context: returning from injury, resuming after a break, switching disciplines
Optional but high-impact inputs
Coach methodology
If you have a specific coaching philosophy or preferred approach, describe it here. The AI Assistant uses this alongside coach memory to shape how it structures the phases — session types, block lengths, progression style, recovery patterns.
Examples of useful methodology notes:
- "Polarised training approach — 80% easy aerobic, 20% high intensity. No moderate-intensity work."
- "Steve Magness philosophy — quality over quantity, short intense blocks rather than high volume."
- "Traditional Lydiard-style periodisation — long aerobic base first, then progressive specificity."
- "Athlete handles 3-week build / 1-week recovery blocks better than 4-week cycles."
If you have used the AI Assistant to generate cycles before, your methodology is likely already captured in coach memory and does not need to be repeated unless you want to change something.
Competitions
List any target competitions within the plan period. For each competition include the name, date, and priority (A, B, or C).
| Priority | What it means |
|---|---|
| A race | Primary goal — the plan is built around it. The AI Assistant will place a specific taper phase leading into it. |
| B race | Secondary goal — the AI Assistant will protect race day but won't restructure the plan around it. |
| C race | Training race — the AI Assistant treats it as a high-quality training session, not a goal event. |
Competitions without dates are ignored. If a competition falls outside the plan period, it does not influence the structure.
How memory shapes the plan
The training cycles generator uses two distinct memories. Both are loaded automatically — you don't need to do anything to activate them.
Coach memory
Coach memory accumulates your periodization style across all generations — how you structure macrocycle lengths, what mesocycle sequences you favour, how you handle tapers, and what coaching philosophy emerges from your notes and methodology inputs.
The more you use cycles generation, the more precisely the AI Assistant reflects your approach. After the first few generations, you will find the initial draft already close to what you would design manually — because the AI Assistant has learned your patterns.
Coach memory is updated after the feedback pass, once the final accepted plan is confirmed. This ensures memory reflects your actual coaching decisions, not just a draft that may have been revised.
Athlete memory
Athlete memory is updated from three sources: workout analyses (session-level observations about the athlete's response, effort calibration, and physiological signals), period analyses (broader training period reviews that capture trends, adaptation patterns, and strategic observations across multiple weeks), and cycles generation (strategic-level context extracted after the feedback pass).
After you accept or refine a cycles plan, the AI Assistant updates athlete memory with what it learned about that athlete from the generation: their current fitness starting point, performance targets, training profile characteristics, and any context from your goals description that should inform future planning.
This means cycles generation improves athlete memory even before any workouts are logged — and the effect compounds: an athlete with 6 months of workout analyses and several cycles generations will have a rich memory profile that shapes every future plan without you needing to restate their background.
The AI Assistant uses athlete memory in cycle planning to make decisions you would otherwise need to state explicitly:
- If the athlete has had recurring calf issues, the AI Assistant will suggest more conservative build phases and explicit recovery mesocycles
- If the athlete adapts well to VO2max work, the AI Assistant may allocate more time to that phase
- If the athlete historically struggles with high-volume blocks, the AI Assistant will build in shorter peak volumes
Athlete memory is not a substitute for stating what's changed. If the athlete has had a recent injury, illness, or significant change in fitness since the last generation, say it explicitly — memory reflects the past, not the present.
What to add explicitly vs. what memory handles
| Let memory handle | State explicitly in inputs |
|---|---|
| Your general coaching philosophy | A change in methodology for this athlete |
| Athlete's typical recovery patterns | Current injury or recent illness |
| Your preferred block lengths | A specific structural approach for this plan |
| Historical phase preferences | A new goal or race that hasn't appeared before |
Memory reflects the past. Anything that has changed recently, or anything specific to this plan that differs from your usual approach, belongs in the inputs — not left to memory.
Reviewing the initial draft
The draft is a plain-text training proposal. Each macrocycle appears as a named section with its date range and a 2–4 sentence description of what it develops. Its mesocycles are listed underneath, indented, each with their own date range and description. A "Periodization Strategy" paragraph at the end summarises the overall logic.
Example structure:
Base Building
25-05-2026 – 05-07-2026 (6 weeks)
Establishes aerobic foundation and running economy before intensity work...
→ Aerobic Foundation
25-05-2026 – 14-06-2026 (3 weeks)
Progressive easy volume, technique work, and functional strength...
→ Strength & Neuromuscular Prep
15-06-2026 – 05-07-2026 (3 weeks)
Short sprints, hill repeats, and explosive strength to prepare...
Each macrocycle will contain at least two mesocycles with distinct training focuses. If you see a macrocycle with only one mesocycle covering the same dates, that's worth questioning in feedback.
Read the draft for:
Structural logic — does the sequence make sense? A typical sound structure moves from aerobic base → threshold development → VO2max/speed → race-specific/taper. Jumps in phase logic (e.g. skipping base entirely for an athlete who hasn't trained in months) are worth questioning.
Date alignment — do macrocycle boundaries land in sensible places relative to competitions? A peak/taper phase should end around an A race. A recovery phase should not overlap with a key race.
Phase length — are the blocks long enough to produce adaptation? A 2-week mesocycle is rarely enough for meaningful physiological change. 3–4 weeks per mesocycle is a minimum for most qualities; 5–6 weeks for aerobic base.
Purpose descriptions — each mesocycle includes a description. Read these for specificity. "Improve fitness" is a weak purpose; "Develop lactate threshold through progressive tempo and cruise intervals" is a good one. Weak purposes produce vague training guidance — specific purpose descriptions are what make a cycle plan actionable.
Feedback recommendations
Feedback is processed as a continuation of the AI Assistant's existing conversation — it has full context of the plan it proposed, so you don't need to restate anything. Just describe what you want to change.
What makes effective feedback
Reference cycle names or phase positions, not just general preferences.
| Less effective | More effective |
|---|---|
| "Make the base phase longer" | "Extend the aerobic base mesocycle by 2 weeks — athlete is coming back from injury and needs a slower build" |
| "Add more speed work" | "The build phase needs a dedicated VO2max mesocycle — split the 6-week build block into 3 weeks threshold + 3 weeks VO2max" |
| "The taper is too short" | "The taper before the A race is only 1 week — extend to 2 weeks and shift the race-specific work earlier" |
| "I don't like the colors" | — not worth using a feedback pass for |
When to accept without changes
Accept the plan if:
- The phase sequence is logically sound for the athlete's goals
- Dates align with competitions
- Mesocycle purposes are specific enough to clearly define what each phase develops
- Block lengths feel appropriate for the training qualities targeted
A plan doesn't need to be perfect to be accepted. Minor preferences about naming or structure are better adjusted manually in GoodCoach after the cycles are created than consumed in a feedback pass.
What feedback cannot fix
- Dates that don't start on Monday — if you want to adjust the start date, it's faster to start a new generation than to request date shifts via feedback
- Missing competitions — if you forgot to include a competition in the initial request, feedback can instruct the AI Assistant to account for it in the structure, but the AI Assistant won't have seen it in its original brief
- Plan period — the total weeks count cannot be changed via feedback; it would require a new generation
Limitations
Minimum 10 weeks
The generator requires at least 10 weeks to build a meaningful periodized plan. Short blocks (4–8 weeks) cannot support the base → build → peak arc that makes periodization effective.
No daily workout detail
Training cycles are periodization architecture, not weekly plans. They tell you what phase each period focuses on — they do not produce daily sessions, volumes, or specific workouts.
No automatic adjustment for future events
Once cycles are saved, they are static records in the Good Coach App. If an athlete's race gets rescheduled or a competition is added, the cycle plan does not update automatically. You would need to manually adjust the cycles in the Good Coach App or generate new cycles after removing current ones.
No constraint validation
Cycles generation does not run safety constraint checks (volume progression, recovery rules, etc.). It operates at the strategic periodization level where those rules don't apply. The quality of the plan depends on the quality of your goals description, methodology inputs, and feedback.
Common mistakes
Vague goals description. "Get faster" gives the AI Assistant nothing to plan around. Specific goals — a target race, a performance metric, a fitness quality to develop — produce specific, coherent plans. The goals description is the most important field; it's worth taking 2–3 minutes to write well.
Skipping competitions. If there's an A race in the plan period and you don't list it, the AI Assistant has no reason to build toward it. The plan will have a generic progression rather than one that peaks at the right moment. Always include competitions when they exist.
Accepting a weak first draft. If the mesocycle purposes are vague ("General training"), the cycles plan gives you nothing actionable to work from. Use the feedback pass to sharpen purpose descriptions before accepting — it costs one extra step but makes every cycle in the plan genuinely useful.
Using feedback for cosmetic changes. Renaming a cycle from "Base Building" to "Foundation Phase", or adjusting a color — these are worth editing manually in the Good Coach App after the plan is saved. Feedback passes are for structural or strategic changes to the plan.
Generating a new plan before the current one runs out. If an athlete already has active macrocycles in the Good Coach App for the target period, generating a new plan may produce overlapping cycles or it will just fail. Either clear the existing cycles first or start the new plan from where the current one ends.