Skip to content

Research and maintenance method

This is a living, question-led evidence platform, not a completed systematic review. Start at the front page; use the glossary before comparing apparently similar training labels.

Reference root and derived copies

docs/ is the authoritative reference root. Topic pages, conclusion records, source records, and paths in this reference, including topics/, evidence/references.json, and topic_file values, are relative to docs/, not to the repository root. Citation aliases and claim IDs are unchanged by that layout. The front page is index.md.

Answers and new-paper updates are maintained through the stable conclusions, their source records, and the owning topic pages. Route a question to an existing conclusion and topic page; do not open a parallel document for a question that page already owns. Revise that conclusion, its sources and the topic page together, using the update workflow.

The repository-root README.md is the GitHub project entry and local publishing instructions. It is not an evidence source. A generated site/ directory, when present, is a derived rendering of this reference; do not edit it or treat it as a second source. initial_prompt.md remains at the repository root and is not part of the published reference.

Scope and provenance

  • Target decision: improve VO₂max, sustainable cycling power, general fitness, and event-relevant performance within approximately 7–12 available training hours per week.
  • Prefer trained amateur cyclists. Elite cyclists, other endurance sports, recreationally active people, and untrained participants must be identified rather than pooled implicitly.
  • Initial evidence check: 2026-10-02. This is a search/access date, not a guarantee that every relevant publication has been found.
  • The starting material is the repository-root file initial_prompt.md, which is not part of this reference. It describes claims from a transcript but does not contain the transcript itself. Papers can be matched to those descriptors; exact quotation, speaker intent, and attribution remain unverified. Published topic scope points question ownership at the question map. Do not send readers to initial_prompt.md to learn which topic owns a question.
  • Initial discovery uses targeted searches, review reference lists, and original-source retrieval. The search register records the queries and access routes actually used. There is no preregistered exhaustive search, independent duplicate extraction, or PRISMA flow; do not describe this as a systematic review.
  • General web search initially failed and was restored during the investigation; the research agents were notified and used it for further discovery and update searches. PubMed search presented an access challenge. Europe PMC records and accessible journal/full-text sources provide direct verification; search-generated summaries are discovery aids, not extraction evidence. Each source records its actual access level.

Authority and document roles

  1. The eight evidence chapters (topics/) are the maintained authority for empirical conclusions. Each important conclusion has a stable claim ID, an evidence grade, population limits, direct references, and uncertainty.
  2. The source register (evidence/references.json) is the bibliographic and extraction record. Its source IDs connect claims to retrieved original records, designs, populations, outcomes, and limitations.
  3. The claim register (evidence/claims.md) routes questions to conclusions. It is an index, not a second independent evidence judgment.
  4. The bibliography (evidence/bibliography.md) exposes the source register in readable form. Its annotations must stay consistent with the extraction record.
  5. The model topic integrates mechanisms and priorities; the program-design topic owns practical application. Example schedules are coaching implementations of supported principles, not experimentally validated packages or proof of an optimum. Starting-claim appraisals and research gaps are linked appendices, not additional independent judgments.
  6. The glossary (CONTEXT.md) owns terminology. Resolve definitions there before silently changing the meaning of a term in a topic brief.

When documents conflict, inspect the underlying studies and update the owning topic page's conclusion first. A newer timestamp alone does not make a statement correct. Reconcile the conclusion register, bibliography, model/program applications and front page in the same update. This reference keeps one live version of each topic page rather than maintaining competing copies.

Evidence hierarchy and extraction

Prefer, in order: systematic reviews/meta-analyses; randomized or controlled interventions; trained-athlete longitudinal studies; consensus/position stands; high-quality observational work; carefully labeled physiological reasoning and expert commentary. Review quality and population relevance can override this rough ordering: a heterogeneous meta-analysis does not automatically outweigh a directly relevant, well-controlled trial.

For every source used for an important conclusion, record:

  • Stable source ID; authors, publication year, title, DOI/PMID when available, and a direct source link.
  • Actual retrieval URL(s), access level (full-text, abstract, or metadata), and checked date.
  • Design; sample size/sex/training history and baseline fitness when available; intervention duration; comparator and outcome.
  • Absolute or relative change and effect size/uncertainty where verified. Within-group improvement is not a between-group advantage. No statistical significance does not prove equivalence.
  • Whether the result is acute, longitudinal, observational, mechanistic, or consensus guidance.
  • Major limitations, adverse effects/recovery costs, conflicting findings, and relevance to the target cyclist.

If a detail is absent from the retrieved source, say “not available in the retrieved abstract/record,” not an invented estimate. Metadata-only access verifies bibliographic identity, not a study's results. Acute signaling, lactate, heart rate, or time above 90% VO₂max cannot on their own establish superior chronic adaptation.

Qualitative conclusion grades

Label Meaning in this platform
Strong evidence Consistent, reasonably robust evidence for the specific claim; directness and population limits still apply.
Moderate evidence Credible support, but limited trials, precision, population transfer, comparator control, or conflicting findings constrain confidence.
Plausible but uncertain Mechanism, acute response, or limited observational/intervention evidence suggests a benefit without establishing the practical claim.
Mostly coaching practice A useful implementation choice with little direct comparative evidence for the exact dose, timing, or optimum.

Do not grade “cycling,” “Zone 2,” or “nutrition” globally. Grade the actual proposition: carbohydrate supports long-session performance may be well supported while 120 g/hour is optimal for every ride is not. These labels are a transparent editorial shorthand, not a formal clinical GRADE assessment.

Updating after a new paper

  1. Route the question. Identify the topic, existing claim IDs, and source IDs before starting a new document. Extend an existing topic unless the subject is genuinely distinct.
  2. Retrieve and verify. Find the original paper; check publication status, corrections/retractions, DOI identity, access level, and publication date. Add explicit recent-date and relevant-population searches, not only author-name searches.
  3. Extract before deciding. Enter the design, participants, intervention, outcomes, magnitude, uncertainty, and limitations in the source register. Track online-first versus issue year consistently; explain a discrepancy rather than guessing.
  4. Compare like with like. Check trained status, baseline fitness, sex, comparator, total dose, diet, test method, and acute versus longitudinal endpoint. Ask whether a new result genuinely contradicts an existing claim or asks a different question.
  5. Revise the existing conclusion. Keep its ID when the question is unchanged. State what changed, supporting and conflicting sources, and whether the grade or practical implication changes. Retain disagreements explicitly; do not overwrite them with an average claim unsupported by any study.
  6. Propagate the decision. Update claim routing, bibliography, affected model/program topic pages, and relevant schedule assumptions together. An exact recipe remains coaching practice unless evidence directly tests that recipe.
  7. Record the update. Add a concise changelog entry with date, affected claim IDs, new sources, changed decision, and remaining uncertainty. A topic page's evidence-check date changes only after a meaningful recheck.
  8. Check navigation and arithmetic. Follow links from index → claim → source; verify local targets/anchors, source IDs, date/DOI consistency, and any affected weekly totals or interval-work calculations.

Answering a future question

A useful answer contains: decision → evidence grade → linked claim → original studies and relevant population → uncertainty → practical consequence. Quote the maintained conclusion, not an uncited recollection of a topic overview. For a question outside the current evidence, state the gap, search it, then update the relevant topic page.

For example, “Are 30/15s better than 4-minute intervals?” requires separating acute time near VO₂max, chronic VO₂max change, test-specific performance, and which long-interval protocol served as comparator. “Can I ride harder on easy days?” requires checking LT1/VT1, weekly dose, hard-session quality, fueling, and the lack of a universal FTP percentage.

Safety and personalization boundaries

These documents are educational programming guidance, not diagnosis or individualized medical care. Persistent unexplained fatigue, dizziness, heat illness symptoms, recurrent illness/injury, disordered eating concerns, menstrual disturbance, or suspected iron deficiency warrant qualified assessment. Do not self-prescribe iron, deliberate dehydration, or heat stress based on a performance claim. Example weeks assume an already-trained, healthy cyclist tolerating similar volume; they are not instructions to jump from inactivity to the listed workload.