AI Endurance Coaching Protocol with Intervals.icu

Purpose and Significance

Section 11 is an open-source protocol that transforms how AI systems deliver endurance coaching. Instead of inconsistent, speculative advice, it provides deterministic, auditable guidance grounded in peer-reviewed sports science. Athletes using ChatGPT, Claude, Gemini, Grok, or Mistral can now receive consistent training recommendations tied directly to their actual performance data—with full reasoning transparency and zero guesswork.

Key Features

• Deterministic Coaching — Same inputs always produce identical outputs, eliminating the randomness that plagues generic AI coaches
• Evidence-Based Framework — Built on 15+ peer-reviewed endurance science models (polarized training, CTL/ATL, HRV, power zones, TSB)
• Full Auditability — Every recommendation traces back to specific athlete data and validated thresholds; no hidden reasoning
• Intervals.icu Integration — Auto-fetches live metrics (CTL, ATL, TSB, HRV, recent activities) via JSON sync, eliminating manual data entry
• Multi-LLM Compatible — Works with ChatGPT, Claude, Gemini, Grok, Mistral, and OpenClaw via standardized prompt protocols
• Structured Output Format — Post-workout analysis includes activity type, power/HR zones, decoupling, variability index, TSS, weekly polarization, and contextual coaching notes
• Pre-Workout Readiness Assessment — HRV, RHR, sleep vs. baseline; TSB and ACWR context; Go/Modify/Skip decision logic
• Athlete-Controlled Data — You own and host your dossier; AI coaches fetch, never store

How It Works

Section 11 operates in three layers: data ingestion, validation, and deterministic output. First, your AI coach fetches your athlete dossier (age, FTP, HR zones, goals, current fitness) and real-time metrics from a JSON endpoint you control (via Intervals.icu export). Next, it validates the data against metric hierarchy rules—readiness indicators (HRV, RHR, feel) take precedence over secondary metrics (TSB, ACWR). Finally, it applies the protocol's logic trees to generate auditable recommendations with no web searches, no speculation, and no contradictions.

Getting Started

Step 1: Create Your Athlete Dossier

Copy the DOSSIER_TEMPLATE.md and populate your profile: age, weight, training goals, current FTP, HR zones, training schedule, and nutrition protocol. This becomes your coaching baseline.

Step 2: Set Up Data Mirror (Recommended)

Export your Intervals.icu data as JSON to a GitHub repository or any web-accessible endpoint. This allows your AI coach to fetch live CTL, ATL, TSB, and recent activities without manual input each session. See the examples/ folder for setup guides.

Step 3: Configure Your AI Platform

Paste the Section 11 instructions into your chosen AI platform's project settings (ChatGPT, Claude, Grok, etc.). Upload SECTION_11.md and your dossier to the knowledge base. Enable web browsing so the AI can fetch your JSON endpoint.

Step 4: Validate with a Test Query

Ask "How was today's workout?" Your coach should automatically fetch your data, analyze it against the protocol, and return a structured report with session metrics, training load context, and a brief interpretation—no asking you for information, no citations.

Who It's For

• Endurance Athletes (cycling, running, triathlon) who want AI coaching that respects their actual data and training science
• Data-Driven Coaches building semi-autonomous coaching tools or athlete management platforms
• AI/ML Developers working on deterministic decision systems, auditable AI outputs, or domain-specific LLM applications
• Teams Using Intervals.icu seeking to layer intelligent analysis on top of existing training data
• Organizations Building AI Agents for sports, wellness, or personalized guidance—Section 11 is a replicable pattern

Core Architecture

The protocol defines three sub-protocols: AI Coach Guidance (11A) covers readiness assessment, workout recommendations, and recovery rules; Training Plan Protocol (11B) handles periodization and load management; Validation Protocol (11C) enforces data source hierarchy and prevents hallucination. Each outputs structured, line-by-line reports with metric priority: Tier 1 (readiness indicators—HRV, RHR, feel), Tier 2 (load-recovery ratios—TSB, ACWR), Tier 3 (diagnostics—decoupling, variability index).

Platform-Specific Setup

Section 11 provides copy-paste instructions for ChatGPT Projects, ChatGPT CustomGPT, Claude Projects, Grok, Mistral Le Chat, Gemini Gems, and OpenClaw. Each setup takes 5 minutes: create a project, paste instructions, upload files, enable web access. The core prompt template is identical; platform-specific configuration is minimal.

Key Resources

• SECTION_11.md — Complete protocol documentation (guidance, validation, output format rules)
• DOSSIER_TEMPLATE.md — Blank athlete profile template
• examples/ — JSON sync setup, report templates, real-world dossier samples
• LICENSE — CC BY-NC 4.0 (free for personal use, attribution required)

Integration with OpenClaw

Section 11 pairs exceptionally well with Antfarm: Multi-Agent Workflow Orchestration for OpenClaw due to its persistent memory and autonomous execution capability. Combined, they enable fully autonomous AI coaching: OpenClaw maintains session context across conversations, automatically fetches updated athlete data, and applies Section 11 logic without user prompting.

Why This Matters

Most AI coaching today is inconsistent—the same question yields different answers, recommendations ignore your actual data, and there's no way to verify the reasoning. Section 11 fixes this by treating AI as a deterministic logic engine, not a creative oracle. It's a blueprint for any domain—medicine, finance, wellness—where auditable, data-driven AI decisions matter more than confident-sounding speculation.