How it works

AI-driven coaching, with a deterministic fallback.

Architecture

What feeds what.

Inputs

01
Your training
Split, days/week, exercise history, load progression.
02
Your check-ins
Sleep, soreness, mood, RPE, readiness signals.
03
Your recovery
Trend lines across weeks vs. your own baseline.
04
Your goals
Where you're headed, your timeline, your constraints.

The coach — source of truth

AI Coaching Layer
Reads every input and writes your plan. Builds your programming, sets your macros, interprets your signals, adapts day to day, and explains its reasoning. Grounded in evidence — it cites what it can support and flags what it can't.

The floor — deterministic fallback

Workout Floor
Exact, repeatable set/rep/progression math. Bounds the AI and runs on its own if the AI is ever offline.
Nutrition Floor
Deterministic TDEE and macro math with hard protein and energy floors the AI can shape but never break.
Readiness Model
Turns your check-ins into a repeatable readiness score that weights today's volume and session cap.

Foundation

The Evidence Base
A structured library of sports-science and physiology research with evidence-quality grading. The AI is required to ground its claims here — and to say so when the evidence is thin.

The coach · AI layer

Reads your week. Writes your day.

The AI is the coach. It reads your training history, your check-ins, and your goals, then builds and continuously adapts your plan — programming, macros, and the daily brief. You can talk to it, ask why, and push back, and it answers in plain language grounded in evidence.

  • Generates and re-generates your sessions from your split, schedule, and recovery
  • Sets and adjusts macro targets to how your body is actually responding
  • Explains its reasoning and cites what it can support

Why this matters

Static templates can't tell the difference between a great week and a wrecked one. The coaching layer reads the difference and rewrites the day around it.

The floor · deterministic engines

Exact math the AI can't override.

Underneath the coaching layer sits a deterministic engine: plain, repeatable math for programming and nutrition. It does two jobs. It bounds the AI — hard floors and safe ranges the AI can shape but never break. And it's the fallback — if the AI is ever unavailable, the engine still produces a sound, conservative plan on its own.

  • Mifflin–St Jeor baseline with activity multiplier and a hard protein floor
  • Deterministic set/rep/progression math, readiness-weighted
  • Runs offline — the plan never disappears because a model is down
Repeatable Bounded Offline-safe

Why this matters

AI is the brain, but a brain needs a skeleton. The deterministic floor is what keeps an off day from turning into a bad recommendation — and keeps the app useful when the network isn't.

The evidence base

Every claim graded. Thin evidence flagged.

The coaching layer is grounded in a structured library of sports-science and physiology research, each entry graded by evidence quality. The AI is required to ground its claims in that library — and when the evidence is limited or theoretical, it tells you instead of pretending otherwise.

strong_human_data limited_human_data emerging preclinical theoretical

What the AI does — and what it doesn't

A clear line. Drawn on purpose.

✓ Does

  • Generates and adapts your workout programming
  • Sets macro targets on the engine's deterministic math
  • Reads your trends and rewrites the day around them
  • Explains its reasoning and cites what it can support
  • Flags claims it can't ground — and stops itself

✗ Never does

  • Makes medical claims or diagnoses
  • Invents numbers the evidence can't support
  • Overrides the deterministic floor
  • Replaces your own judgment or your coach's

Educational and coaching tool only. Not a substitute for professional medical advice. Always consult a qualified professional before making significant changes to your training, nutrition, or health.