Exercise definitions
Stable movement facts: name, muscles, equipment, movement pattern, and logging compatibility. These seed future prescriptions but should not rewrite an active routine.
The RAMM System
How RAMM turns training logs into explainable next-session goals.
RAMM is designed around a traceable training model rather than a black-box workout generator. The system separates the planned prescription, the performed workout, and the progression decision so future goals can be explained instead of inferred from the last number logged. Some safeguards described below represent RAMM’s product direction and may not yet be exposed in every part of the current app.
Core idea
A workout result is evidence, not the next prescription.
Many workout apps collapse planning and logging into one loop: you perform a set, the app copies the result forward, and next week’s target drifts with whatever happened that day. That can be convenient, but it can also make the next target reflect a noisy training day rather than the intended progression rule.
If you planned 28 kg × 11 / 11 / 11 and performed 28 kg × 11 / 8 / 11 / 7, the weak sets might mean fatigue, poor setup, a rushed rest period, or a bad target. RAMM’s direction is to preserve the performed result, capture context, and let a deterministic decision decide what can safely carry forward.
In this example, the fourth set was added during execution. It remains part of the workout record, but it does not automatically turn the underlying three-set prescription into a four-set prescription.
Fail closed when data is incomplete.
If a routine lacks structured prescription data, or a substitution cannot be compared safely, goal generation should require review instead of falling back to generic defaults.
Progression profiles
Rules and constraints for how a movement can progress. A dumbbell press, pull-up, cable isolation, and core drill should not all progress the same way.
Routine prescriptions
The exact sets, ranges, role, goal type, and progression profile chosen for a routine. This is where the plan becomes user-specific.
Weekly goal plans
Activated goals for a training week. Once active, they are treated as an audit trail, not a mutable scratchpad.
Workout execution
What actually happened in the gym: performed loads, reps, substitutions, skipped sets, and intent such as fatigue or controlled lower-volume work.
Progression decisions
The deterministic interpretation of the completed work: increase, hold, reduce, or require review when the evidence is ambiguous.
Recovery context
Intended supporting inputs such as fatigue or readiness context. These can help explain caution, but should not silently rewrite history or invent prescription values.
Workflow
From routine creation to the next goal
RAMM’s useful distinction is that a routine, a workout session, and a weekly plan are not the same object. A routine describes what should be trained. A workout session records what happened. A weekly plan activates goals for a period of time.
- 1A routine is published with structured prescriptions.
- 2A weekly plan activates targets for the current training week.
- 3The workout screen shows the planned target and the matching previous set context.
- 4The lifter logs what happened, including any set intention or substitution.
- 5RAMM’s goal architecture is designed to compare the completed session with a preserved prescription snapshot.
- 6A progression decision is produced: progress, hold, reduce, or review required.
- 7The next weekly goal plan is generated from that decision trail.
Boundaries
What can safely change?
- Canonical exercise definitions can improve future routine creation.
- Published routine prescriptions should not silently re-read changed canonical defaults.
- Activated weekly plans should be versioned instead of edited in place.
- Workout execution can record temporary changes without making them permanent.
- AI can explain an existing decision, but should not invent prescription values.
Worked example
Example: from prescription to next-session goal
1. Routine prescription
A routine prescribes three dumbbell-press sets at 28 kg for 11 repetitions. The prescription belongs to the routine structure: exercise, set count, target range, role, and progression profile.
2. Activated weekly goal
The weekly plan activates that target for the current training week without changing the routine itself. The goal is now a versioned training target, not a live copy of future canonical defaults.
3. Workout execution
During the workout, the lifter records 11, 8, and 11 repetitions, then adds a fourth set of 7. The session preserves all four performed sets and the reason for the added work.
4. Progression decision
The progression layer evaluates the result against the original three-set prescription and the exercise’s progression profile. It does not automatically reduce every future set to eight reps or permanently add a fourth set.
5. Next goal or review
If the evidence is clear, the system can hold, progress, or reduce the next target. If the added set or execution context makes the comparison ambiguous, the decision requires review. A later weekly plan can activate the resulting goal as a new version while retaining the earlier plan as history.
When actual performance differs from the plan
Deviations are normal. You may reduce load because sleep was poor, increase reps because the target was too easy, or swap an exercise because equipment was unavailable. The important question is not “what number did you log?” but “what does this result mean for the next decision?”
RAMM’s execution flow is designed to keep that distinction visible. The performed result stays in the log. Intent metadata explains why the result changed. The future goal is decided later against the routine prescription and progression rules.
Learn more in Workout Execution.
Where AI belongs
RAMM’s public direction is intentionally conservative: AI should not invent weights, reps, set counts, increments, measurement types, or progression profiles. Those values belong in structured data and deterministic rules.
AI becomes useful after the decision exists. It can explain why a goal held steady, summarize the reason a review was required, or translate a deterministic decision into plain coaching language.
Read the boundary in AI Insights.
Frequently asked questions
How RAMM handles training decisions
Does RAMM automatically increase weight every time?
No. A completed workout is evidence, not a command. RAMM should increase, hold, reduce, or require review depending on the prescription, performed sets, progression profile, and quality gates.
Can a completed workout change an active weekly plan?
Activated weekly plans should remain immutable. If goals need to change, RAMM should create a new version rather than silently editing the historical plan.
What happens when I train lighter than planned?
The performed result is logged, and the reason matters. A lighter set caused by fatigue, equipment availability, technique constraints, or a deliberate lower-intensity choice should not always produce the same next goal.
Does AI choose my weights and reps?
No. RAMM’s product direction keeps prescription values inside structured data and deterministic rules. AI may explain an existing decision in clearer language.
What does review required mean?
It means the system should fail closed instead of guessing. Missing history, incompatible substitutions, unclear set counts, or ambiguous execution can require human review before activation.
Explore the system next
Ready to train with a plan?
Join the early access list and be the first to experience RAMM.
No spam. Just early access updates.