Weekly Check-In Triage: 3 Signals That Beat AI Guesswork

Weekly Check-In Triage: 3 Signals That Beat AI Guesswork

When coaching decisions are made from one week of noisy data, the best systems don’t predict everything—they sort what matters first.

The strongest finding in the Kahunas coaching corpus is not that one metric predicts everything; it’s that fast, situation-specific triage improves decisions when the signal is messy. Justin’s repeated pattern is plain: check the recent trend, identify the mechanism, then make a concrete call—adjust insulin when the same dose starts pushing meals earlier, pause or reduce retatrutide when appetite suppression starts fighting the next phase of the plan, and hold or drop growth hormone when blood sugar worsens. That is the mechanism: decision triage under uncertainty. The falsifiable thesis is simple: weekly check-ins work best when AI is used to rank the biggest constraint first, not when it tries to optimize every variable at once.

Check-ins are not reports; they are triage

Most coaching software treats the weekly check-in like a form to be completed. That’s the wrong mental model. In the examples we actually have, the value comes from sorting the week into a small number of decisions:

1. What changed?
2. What is the mechanism?
3. What should change now?

That sequence matters because body composition work rarely fails from a single dramatic event. It fails from small mismatches that keep compounding: appetite no longer matching calorie targets, insulin no longer matching meal timing, recovery no longer matching workload, or blood glucose no longer matching the current drug/support setup. The coach who sees those mismatches early wins the week.

Joe Webb’s check-in is a clean example. He reports that on his high day, the same insulin dose that worked last week now dips his blood sugar noticeably, so he has to bring meal 2 forward by about 30 minutes. He doesn’t binge, he doesn’t panic; he simply reports the new constraint and says he’ll reduce the dose further on the next high day. Justin’s response pattern in this kind of scenario is not philosophical. It’s operational. When the same dose starts acting differently, the dose is no longer a fixed truth. It is a moving input.

That is the first coaching lesson for AI systems: the useful output is not a prediction; it is a ranked action list.

Signal 1: timing drift beats vague “progress” language

The best weekly check-ins often expose timing drift before they expose weight drift. That’s useful because timing drift is easier to act on quickly.

In Joe’s case, the issue wasn’t just “insulin sensitivity improved.” The more useful detail was that meal timing had to shift forward to keep the day on track. That tells the coach the system is still functioning, but the margin has changed. The decision quality comes from noticing that the problem is local, not global.

AI check-in triage should therefore ask for and rank:

• meal timing changes
• appetite changes around specific meals
• training performance changes tied to a day type
• any new need to compress or expand the schedule

If an AI assistant buries that beneath a long summary of “overall adherence,” it misses the actual decision point. A coach does not need six paragraphs to know that a high day now needs a dose adjustment. The software should surface the altered constraint immediately.

This is where a lot of hype goes wrong. People talk about AI as if its job is to replace coaching intuition. In practice, the more reliable role is much narrower: force the weekly data into a decision order. If the check-in says, “same dose, earlier meal, no overeating,” the answer is not “interesting.” The answer is, “the dose is too high for the current response; reduce and re-test.”

Signal 2: appetite suppression is only useful if it matches the phase

Rory Lazowski’s retatrutide notes show a different kind of triage problem. He reports 2 mg producing almost no appetite, even on low-carb days, and mentions more fatigue than normal. He also says he may pause or reduce the dose if food increases later, while noting that it could be useful for gaining at smaller doses.

Justin’s response is revealing because he does not treat “appetite lower” as automatically better. He says he’s trying it himself to build firsthand experience for clients, but he’s not fully convinced by the idea of forcing appetite lower. That skepticism matters. In bodybuilding coaching, appetite suppression can solve one problem and create another. It can make a cut easier while making a gaining phase harder if food intake has to rise later.

So the check-in triage question is not “did the compound work?” It is:

• What phase are we in?
• Does this appetite effect help the current phase?
• Does it create a bottleneck for the next phase?

That is a better decision framework for AI than raw sentiment analysis or “wellness scores.” The same intervention can be useful in one phase and a nuisance in another. A good weekly triage system should therefore tag appetite-related changes by phase, not treat them as universal wins.

Justin’s call in the Rory exchange is also practical: if the body composition situation makes leaning out easier, use it while it’s easy. That is not an emotional response. It is a timing response. When a tool removes friction from the current phase, the coach can exploit that window rather than wait for the perfect future state.

Signal 3: blood markers matter when they change the plan, not when they decorate it

The clearest example of mechanism-first triage in the KB is the growth hormone discussion in the coaching exemplars. Justin frames it as a real tradeoff: if the only question is bodybuilding, GH wins even if it forces more insulin. But if it worsens blood sugar, it is working against the health issue being addressed. He also notes that higher year-round doses push most people onto the wrong side of the curve and that the point of diminishing returns is different for everyone and unpredictable.

That is the part many weekly check-ins never say out loud: not every data point deserves equal power. A number matters when it changes the plan. If blood sugar worsens enough to alter the usefulness of the current support setup, that is a decision point. If it doesn’t, it is just noise.

For AI coaching, this means a check-in system should not just record markers; it should compare markers to the last decision threshold. Did the current support still do its job? Did the tradeoff move? Did the athlete cross from “useful” to “too costly”? If so, the next step is not “monitor.” The next step is change.

That’s the skeptical part of the argument: more data does not automatically improve decision quality. More data only helps when the system knows which changes are reversible, which are phase-dependent, and which demand immediate action. Otherwise, AI becomes a very expensive filing cabinet.

What a better weekly triage flow looks like

A useful check-in workflow for coaches should be brutally simple:

1. Identify the primary bottleneck. Is it appetite, meal timing, energy, blood sugar, workload, or adherence?

2. Attach the bottleneck to the phase. Does this help the current goal or obstruct it later?

3. Make the smallest viable change. Reduce a dose, shift meal timing, pause a support tool, or hold steady if the system is still inside tolerance.

4. Re-test on the next check-in. The point is not certainty. The point is cleaner information next week.

That framework fits the real examples better than broad “optimization” language. Joe’s insulin issue is a timing adjustment. Rory’s retatrutide is a phase-fit question. The GH example is a tradeoff question. Those are different decisions, and a good AI system should not flatten them into one generic score.

The practical standard for coaches

If AI coaching wants credibility, weekly check-ins need to answer one question first: what decision should change because of this data?

That standard is deliberately narrow. It does not ask AI to be creative. It asks AI to be disciplined.

When a coach reads a check-in and immediately sees that the same insulin dose now shortens the time to the next meal, or that appetite suppression now helps the cut but may block the gain phase, or that a support choice is no longer worth the tradeoff, the system is doing its job. If the software can’t surface that kind of triage, it is producing content, not coaching.

That’s the thesis worth testing in real practice: weekly check-ins improve decision quality only when the AI ranks the constraint, names the mechanism, and recommends the smallest next move. If it can’t do that, the athlete gets more text but not better coaching.

Sources Used

• raw/_consumed/2026-05-26/troponiniq_kb.md
• raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w13-18m/clients/rory_lazowski___members-c5balaovjbdoeefqmfuqdhh2tbpmfdu16lnf0tnrtmw.json
• raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w13-18m/transcripts/rory_lazowski___members-c5balaovjbdoeefqmfuqdhh2tbpmfdu16lnf0tnrtmw.md
• raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w13-18m/clients/joe_webb___members-rksigkykimaxwmo_t4_e8nwvbtc2j0etleutkyysads.json
• modules/03-knowledge/kahunas-coaching-deep-nutrition.md
• modules/08-voice/kahunas-coaching-deep-voice.md