Weekly Check-In Triage: 3 Signals from Justin Harris’s Coaching Logic

Weekly Check-In Triage: 3 Signals from Justin Harris’s Coaching Logic

A practical decision-quality framework for AI fitness coaching, built around what to ask, what to ignore, and when to escalate

The strongest recurring coaching move in the Kahunas corpus is not a better answer — it is a better triage: reduce noise, identify the one signal that changes the plan, and avoid inventing precision where the data do not support it. In the contest-prep cases, Justin Harris explicitly rejects body-fat guessing because calipers are wrong and even hydrostatic testing is only an approximation; in the same thread, the real decision is whether the physique is on track to fill out and refine over the next month. That is a mechanism of decision quality: signal suppression. The thesis is simple and falsifiable: weekly AI check-ins improve coaching only when they help a coach classify the athlete into one of a few action buckets faster than a human would by free-form reading.

Why weekly check-ins are a triage problem, not a content problem

Most coaching systems treat the weekly check-in like a miniature consultation. That is the wrong frame. In practice, the value of a weekly update is not that it contains every detail; it is that it reveals whether the current plan should continue, tighten, or be escalated.

The Kahunas coaching notes point to this repeatedly. In the diet-periodization material, the core rule is that most results come from nailing macros; fruit versus pasta is a marginal issue unless the athlete is already close to the limit of what the plan can absorb. In other words, the check-in should not be optimized for culinary correctness. It should be optimized for whether the athlete is still inside the operating range.

That is exactly where AI can help if it is used as a triage layer. A weekly check-in can be structured to answer three questions:

1. Is the athlete executing the current plan?
2. Are the trend metrics moving in the intended direction?
3. Is there a red flag that changes the decision category?

If the system cannot separate those three, it is just producing more text.

The first job: classify execution before you interpret outcomes

A coach cannot evaluate outcomes without knowing whether the plan was actually followed. This sounds obvious, but it is the first place AI check-ins go soft. They over-read bodyweight fluctuations, training pump, hunger, digestion, and mood before they establish whether the athlete missed meals, drifted carbs, changed food choices, or changed activity.

The nutrition cases in the KB keep returning to the same hierarchy: the macro targets drive the result, and the smaller food-choice details matter mainly after execution is already stable. Justin’s reasoning on fruit is a useful example. Fruit can be fine pre- and post-workout up to about half the carbs in those meals, and on high days sugar matters less because carbs are high and insulin is elevated all day. But that guidance is conditional on the larger structure being intact. If the athlete is already off-plan, debating banana versus blueberries is premature.

For AI, that means the check-in should begin with execution flags, not with interpretation. A useful triage prompt is not “How did the week go?” but:

• Were all prescribed meals and macros hit?
• Were any meals substituted?
• Were training sessions completed as written?
• Was cardio, steps, or activity changed?
• Was there any unplanned refeed, binge, travel, or schedule disruption?

This is boring on purpose. Boring is what makes triage work.

The second job: read direction, not isolated numbers

Once execution is known, the question becomes whether the trend matches the plan. The contest-prep material is blunt here: photos matter more than abstract estimates, and the real question is whether the athlete is moving toward the intended look. In one case, the athlete is already very lean, single digit, and the practical issue is not a body-fat number but the month or so needed to fill out and thin the skin for stage presentation.

That is the right way to think about weekly check-ins in general. A single weigh-in is not a verdict. A single pump is not a verdict. A single bad sleep night is not a verdict. The check-in is supposed to detect a direction of travel.

So an AI coaching system should force comparison against prior weeks:

• Is bodyweight moving at the expected rate for the phase?
• Are waist and visual markers consistent with the phase?
• Are hunger, recovery, and training performance trending in a way that matches the current deficit, maintenance, or surplus?
• Is the athlete’s appearance changing in the expected direction, or is the plan producing the wrong adaptation?

This matters because decision quality depends on relative change, not raw description. A lean athlete who looks a little flatter after a few days of hard dieting may simply need more time to fill back out. A heavier athlete who is losing scale weight but not improving visually may need the plan reconsidered. The system should not make those calls from a single datapoint.

The third job: identify escalation triggers early

A triage system is only useful if it knows when not to keep the issue in the normal weekly loop. The deep PEDS cases show this logic clearly: when growth hormone worsens blood glucose, the question is no longer about elegance or preference. It becomes a tradeoff between bodybuilding benefit and a worsening signal. Justin’s answer is not “optimize around it forever”; it is that there is a point where the tradeoff changes and the current setup stops being the best decision.

You do not need to import the drug discussion into regular coaching to use the logic. The general principle is: some check-in inputs are not routine noise. They are escalation signals.

Examples of escalation signals in a fitness coaching context include:

• repeated inability to hit the plan despite intent,
• a clear mismatch between expected and observed trend,
• persistent digestion or food tolerance issues that make compliance impossible,
• performance collapse that is not explained by the phase,
• or a physique result that is moving the athlete away from the goal despite adequate adherence.

The key is that escalation is not failure. It is a category change. Good weekly triage tells the coach when the normal adjustment cadence is no longer enough.

What AI should actually do in the weekly check-in

The useful AI role is not “analyze everything.” It is to compress the week into a decision memo.

A strong check-in triage template does four things:

1. Summarizes execution in one line. Did the athlete follow the plan closely, partially, or poorly?

2. Extracts trend direction. Are bodyweight, photos, performance, digestion, hunger, and adherence moving in the intended direction?

3. Labels the current state. On track, slightly off, materially off, or requires escalation.

4. Recommends the next decision. Hold, make a small adjustment, tighten execution, or intervene.

That is more valuable than an AI that writes a paragraph about “overall progress” while failing to identify whether the athlete actually adhered.

Why this is better than more “intelligence”

The temptation in AI coaching is to imagine that more context automatically produces better decisions. The KB argues the opposite. Better decisions often come from refusing low-value precision and concentrating on the few variables that change action.

That is why the body-fat guessing case matters. A number can feel precise and still add no useful information. It is also why carb-source debates matter less than total compliance. And it is why a weekly check-in should be designed as a funnel, not an essay.

If the athlete is on track, the right output is usually a small or no change. If the athlete is off track for a reason that can be corrected with a small tweak, the right output is a precise tweak. If the athlete is off track because the plan itself no longer fits the situation, the right output is escalation. The system earns its keep by sorting those cases faster and more reliably.

The practical standard for coaches

A weekly check-in triage system is good if it improves one thing: the odds that the coach makes the right decision on Monday from data collected over the week.

Not the most verbose decision. Not the most confident-sounding decision. The right one.

If you want a clean test for any AI coaching workflow, use this: after reading the check-in, can the coach name the current category without re-reading the whole message? Can they say whether to hold, adjust, or escalate? If yes, the system is helping. If no, it is generating documentation instead of decision support.

That is the standard worth building toward, because weekly coaching only works when triage is better than intuition — and better than noise.

Sources Used

• modules/03-knowledge/kahunas-coaching-deep-nutrition.md
• modules/03-knowledge/kahunas-coaching-deep-contest-prep-peaking.md
• modules/03-knowledge/kahunas-coaching-deep-peds.md