The Weekly Check-In Triage Rule of 3 Signals

The Weekly Check-In Triage Rule of 3 Signals

A practical decision filter for AI fitness coaching: what to trust, what to ignore, and what to escalate

Justin Harris’s coaching notes are blunt on the main failure mode: if the scale is moving up over short periods, it is mostly water or fat, and water shifts can even support rebound-related myofibrillar growth; in the recovery-tracking guidance, the recommended replacements are training performance, energy quality, sleep, mood, and recovery rate. That is the mechanism in one phrase: fluid-shift noise masking actual adaptation. The falsifiable thesis is simple: weekly check-ins get better when AI triage de-prioritizes scale noise and routes decisions through performance and recovery signals first; if you keep promoting the scale to the top slot, you will make worse coaching calls.

Weekly check-ins are not data dumps

A check-in is not valuable because it has more inputs. It is valuable because it improves the next decision.

That distinction matters in AI coaching, where software can ingest far more than a human coach can. If the system treats every metric as equal, it becomes a very efficient way to produce confusion. The practical job is triage: decide what deserves action this week, what deserves monitoring, and what should be ignored for now.

The KB sources here point to a clear ordering. First, the scale is a lagging, noisy measure during metabolic recovery. It captures body weight, not body fat, and early jumps can reflect glycogen, water, digestive content, and hormonal fluid shifts. The recovery guidance explicitly calls out a 3–7 lb increase in the first few weeks of reverse dieting as restoration, not regression. That is not a reason to stop tracking scale weight; it is a reason to stop letting it drive the decision.

Second, the better check-in inputs are the ones that move before body composition does: training performance, energy, sleep, mood, and how quickly someone recovers between sessions. Those are the fields a coach can use to infer whether the current plan is working or whether the athlete is accumulating fatigue faster than adaptation.

Third, check-in triage becomes much more reliable when it asks one question before any adjustment: is this a signal or just noise?

The order of operations

A weekly AI coach should sort check-ins in this sequence:

1. Safety and disruption. Did the athlete report a major issue that changes the plan today? If yes, pause and route appropriately.
2. Training performance. Are the lifts going up, stable, or clearly sliding?
3. Recovery state. How are sleep, mood, energy, soreness, and readiness trending?
4. Nutrition execution. Were targets actually followed, or did the week simply not match the plan?
5. Scale trend. Only after the above.

That order is not cosmetic. It prevents the most common bad call: changing calories because the weight jumped while the actual adaptation markers improved.

Justin Harris’s notes on offseason weight targets make the same point from another angle. He says the scale can be pushed to a desired number quickly, but that forcing a particular weight is usually a net negative to progress and leaves the athlete worse off by year-end. The mechanism is obvious enough: if the number can be moved faster than tissue can be built, then the number is not the thing you are trying to optimize.

For coaches building AI systems, that means the model should not be asked, “What should we change?” before it answers, “What changed for real?”

A practical triage rubric for weekly check-ins

Here is a version that is actually usable.

Green: keep the plan

Use green when:

• performance is stable or improving,
• sleep and mood are acceptable,
• recovery between sessions is normal,
• adherence is reasonable,
• scale changes are explainable by the current phase.

In a green week, the coach does not need to get clever. The system should produce a short summary and a no-change recommendation. This is where AI can help most: normalize the reporting, spot trends, and stop the coach from overreacting to a single noisy datapoint.

Yellow: gather one more week

Use yellow when the picture is mixed.

Common cases:

• the scale is up, but training is better and recovery is fine,
• the scale is flat, but performance is rising and the athlete is not depleted,
• the scale is down, but sleep, mood, or training output are deteriorating,
• adherence is unclear.

Yellow is not a shrug. It is a structured delay. The right move is often to hold the plan for another week, ask for better reporting, or clarify whether the issue is execution rather than programming.

That matters because many “plateaus” are actually bad classification. The athlete thinks they are stalled; the coach sees a number and panics; the real issue is that the inputs were never good enough to justify an adjustment.

Red: adjust now

Use red when the check-in shows a genuine problem that is consistent across signals.

Examples:

• performance is dropping,
• recovery is clearly worsening,
• fatigue is accumulating,
• adherence is good but the current setup is no longer tolerated.

In red, the point is not to chase a better mood. It is to make the smallest change that addresses the underlying problem. AI is useful here if it keeps the recommendation narrow. Big swings are often a sign that the system failed the triage step.

Why this matters more in AI coaching than in manual coaching

A good human coach already does some of this intuitively. The risk with AI is that it can be too literal.

If a check-in template includes body weight, waist, steps, sleep, stress, hunger, recovery, soreness, and workout performance, the model can generate a confident narrative around almost any combination. Confidence is not decision quality.

The check-in triage rule should therefore be conservative:

• do not let one noisy measure override multiple stable ones,
• do not let a better-looking number hide declining performance,
• do not make plan changes when the athlete is merely restoring normal hydration and glycogen,
• do not reward scale movement if the underlying markers are heading the wrong way.

This is where the recovery guidance and the coaching notes line up cleanly. Recovery is not the same as fat gain, and bodyweight gain is not the same as productive progress. The coach’s job is to identify which of those is happening before the athlete starts making decisions from the wrong signal.

The simplest test for weekly decision quality

If you want to audit an AI coach, ask whether it can answer these three questions after every check-in:

• What is the best evidence that the plan is working?
• What is the best evidence that the plan is not working?
• What is the smallest decision that matches the evidence?

If it starts with the scale, it is probably too eager.

If it starts with training performance and recovery, and only then checks the scale for context, it is behaving like a coach rather than a dashboard.

That is the whole point of triage. Weekly check-ins should not be about collecting more numbers. They should be about making fewer, better decisions.

Sources Used

• raw/kahunas-export/2026-05-28/clients/michael_main___members-a2m88q4kyryqrsbdgta-x0mipybv-fzeobfolztzovk.json
• wiki/drive-nutrition-recovery-tracking-and-biofeedback.md