Weekly Check-In Triage: 3 Signals That Improve AI Coaching Decisions

Weekly Check-In Triage: 3 Signals That Improve AI Coaching Decisions

Decision quality beats response volume when the check-in is the control room, not the scoreboard.

The clearest check-in signal in the KB is simple: Joe Webb noticed that the same insulin dose started driving his blood sugar lower enough that he had to move meal 2 up by about 30 minutes, and Justin Harris responded by reducing the dose further on the next high day. That is a mechanism-level example of real-time triage: sensitivity changed, the plan changed. The thesis is falsifiable: weekly AI coaching works best when it routes the athlete to the next decision by tracking a few actionable signals, not when it tries to narrate everything.

The weekly check-in is not a recap

A lot of coaches treat the check-in like a diary review. That’s a mistake. In the materials here, the best decisions come from separating three questions:

1. What changed?
2. Is the change useful, neutral, or a problem?
3. What is the smallest adjustment that preserves momentum?

That order matters because it prevents two common failures: overreacting to noise and waiting too long to correct a real shift. The Joe Webb example is the cleanest version of this. The client did not report a vague feeling. He reported a concrete operational change: the same insulin dose lowered blood sugar more than expected, forcing meal timing to shift. That is not just data; it is a decision trigger. The coach did not ask for more commentary. He reduced the dose further. The logic is straightforward: if the same dose now has a bigger effect, the dose is too high for the current state.

That is the kind of triage AI coaching should copy.

Signal 1: the body’s response has changed

The first triage signal is any change in response to a standard input. In the KB, this appears in several forms, but the principle is the same: if the same lever produces a different outcome, the athlete’s current state has changed.

You can see it in the Joe check-in with insulin timing. You can also see the same decision style in Justin’s voice example on growth hormone: he treats worsening blood glucose as a real tradeoff, notes that higher year-round doses push many people onto the downward slope of diminishing returns, and says the current dose is likely too much given reduced sensitivity. The plan call is not complicated: drop it for now and watch the numbers over a month or two.

The mechanism is not magic. It is state dependence. Appetite, glucose handling, fatigue, food tolerance, and recovery do not stay fixed while the plan remains constant. If an AI coach does not anchor on state change, it will keep recommending yesterday’s answer to today’s physiology.

For weekly triage, this means the first pass should be built around response drift:

• same food, different hunger
• same dose, different effect
• same volume, different digestion
• same training load, different recovery

If the athlete can describe the drift in plain terms, the coach can usually decide quickly whether to hold, reduce, or add.

Signal 2: the change affects execution

Not every change matters. The second triage signal is whether the change disrupts the plan.

In Joe Webb’s case, the altered insulin response did. He did not overeat, but the meal schedule had to compress. That matters because bodybuilding prep is not just about average intake; it is about how reliably the intake can be executed. A plan that is technically correct but keeps forcing compensations becomes fragile.

Justin’s retatrutide commentary shows the same logic from the appetite side. The effect he reported was very strong appetite suppression after 2 mg, plus a bit more fatigue than normal. He was clear that he was unsure he liked the idea of forcing appetite lower, and he was hesitant about using it as food increases later in the year. The practical decision was to run with it while leaning out, because that matches the current objective and gives a better example of how it behaves in a gaining phase later.

That is a useful triage pattern: when a tool makes execution easier for the current phase, keep it. When it makes future phases harder to interpret, be cautious. The coach is not trying to optimize every variable at once. He is trying to keep the plan executable in the phase that matters now.

This is especially relevant for AI systems because they can overfit to the loudest signal. If appetite is suppressed, the machine may call that success. If food volume is low and fatigue is up, it may call that fine. But a coach has to ask a sharper question: does this help the athlete complete the week as designed?

Signal 3: the change alters the next phase’s decision quality

The third signal is more subtle and more important: does this week’s decision improve the quality of next week’s decision?

That question shows up in the food-volume discussion from Justin’s podcast comments. He is not obsessed with adding muscle off season as a standalone metric. He frames the off-season around teaching the body to digest and assimilate a large amount of clean food, reaching the point where the athlete can eat more without gaining weight, and doing it gradually rather than in two days. The reason is decision quality later: the better the athlete handles food now, the better the metabolic and prep outcomes later, and the more muscle is preserved in contest prep.

Whether or not you agree with every framing choice, the coaching logic is consistent. Build the state that makes future decisions easier.

That is exactly why weekly check-ins should not only ask “what happened?” They should ask:

• Did this week improve food handling?
• Did this week clarify appetite behavior?
• Did this week reveal whether the current dose is too much, too little, or just right?
• Did this week reduce uncertainty for the next phase?

If a check-in gives you data but no decision advantage, it is low quality. If it gives you a smaller, cleaner decision next week, it is high quality.

What AI should do in the check-in triage layer

A lot of AI coaching discussion starts too high, with personalization or empathy or “24/7 availability.” Those are features. The real test is whether the system can help a coach triage an athlete’s week without wasting time.

The KB points toward a practical workflow:

1) Detect state change

Read for deviations from baseline, not just adherence. A standard dose hitting differently, appetite changing sharply, fatigue creeping up, or meals needing timing changes are the kinds of signals that matter.

2) Classify the impact

Ask whether the deviation improves execution, harms execution, or is irrelevant noise. Retatrutide lowering appetite may be a win during a cut and a headache during a gaining phase. That distinction changes the recommendation.

3) Make the smallest useful adjustment

Justin’s responses are rarely dramatic. Reduce the dose. Hold and watch. Run with it while the goal matches the effect. That style is not indecision; it is precision. The adjustment is only as big as the change requires.

4) Set the next observation window

The purpose of triage is not final judgment. It is to determine what should be monitored next week. If the dose changes, the coach needs the next check-in to answer whether the response stabilized. If appetite drops, the next check-in needs to show whether intake still matches the plan.

The practical standard for weekly AI coaching

If you want a high-quality weekly check-in, make the AI answer three questions before it offers advice:

• What is different from last week?
• Does that difference help or hurt the current phase?
• What is the smallest change that preserves the larger plan?

That is more useful than a long summary and more defensible than blanket personalization claims. It also matches the actual coaching pattern in the KB: state change, execution impact, minimal adjustment.

AI coaching gets better when it becomes a triage layer, not a speech generator. The point is not to comment on everything the athlete reported. The point is to identify the one or two changes that alter the decision. When the same input starts hitting differently, change the plan. When a tool helps the current phase, keep it. When it damages future decision quality, slow down. That is the weekly check-in standard worth building.

Sources Used

• raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w13-18m/clients/joe_webb___members-rksigkykimaxwmo_t4_e8nwvbtc2j0etleutkyysads.json
• raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w13-18m/transcripts/rory_lazowski___members-c5balaovjbdoeefqmfuqdhh2tbpmfdu16lnf0tnrtmw.md
• modules/03-knowledge/kahunas-coaching-deep-nutrition.md
• modules/08-voice/kahunas-coaching-deep-voice.md
• raw/Justin_TT1.txt