Meal Timing and the 1-Week Rule in AI Nutrition Coaching

Why small nutrition edits should follow trend data, not every check-in ping

Joe Webb reduced insulin by 1 IU on his high day, then had to move meal 2 up by 30 minutes; when he repeated that lower dose with meal 3, the same thing happened again. That is the mechanism in plain terms: improved sensitivity changes nutrient handling fast enough to alter meal timing before bodyweight or physique has time to catch up. In a coaching system that updates too often, you end up reacting to noise; the falsifiable thesis is simple: nutrition changes should be made on a trend window, not on each new data point.

AI coaching makes over-adjustment easier than it is useful. The machine can surface a fresh weigh-in, a different appetite report, a slightly flatter pump, or one messy meal and instantly recommend a tighter carb split or a new high-day rule. But the case data in the Kahunas corpus points in the opposite direction: the useful change is usually the one that preserves the plan long enough to see whether the signal is real. The best evidence for that is not a grand theory. It is the practical fact that when Joe noticed better insulin sensitivity, he did not overeat for the day; he just brought meals closer together and planned to reduce the dose further on the next high day. That is a timing adjustment, not a plan rewrite.

That distinction matters because timing effects show up sooner than outcome effects. If a client’s insulin sensitivity improves, meal timing can need to shift inside the week. If a client starts a compound that strongly suppresses appetite, appetite may change within days. Justin’s own retatrutide note is a good example: 2 mg on Friday led to “no appetite whatsoever” through Saturday and Sunday, with more fatigue than normal. In the same exchange, he did not jump to a sweeping conclusion about gaining or cutting; he said he would give the retatrutide a pause or at least reduce the dose if food was being added. The practical point is not that appetite suppression is universally good or bad. The point is that when a variable acts quickly on intake, the coach should respond quickly only if the downstream target has also changed.

That “if” is where a lot of AI coaching goes wrong. It confuses a changed intermediate variable with a changed objective. Appetite is an intermediate variable. Insulin sensitivity is an intermediate variable. Food volume tolerance is an intermediate variable. None of them is the whole outcome. In the Rory Lazowski exchange, Justin was skeptical of the idea that retatrutide is automatically “helpful in gaining,” because he had not yet reconciled that with the reality that people can eat less even when they need more food. He held the opinion until he had more data. That is the right level of skepticism for coaching systems: don’t declare a mechanism useful just because it is potent.

The daily lesson for nutrition timing is that you should separate fast signals from slow decisions. Fast signals include appetite, meal spacing, and whether a given insulin dose now nudges glucose lower than it did last week. Slow decisions include whether the overall macro structure still fits the phase, whether the athlete is actually recomping, and whether a supplement or dose belongs in the off-season, prep, or neither. Justin’s own practical line on the Rory thread was blunt: with the body comp as it was, run with the retatrutide and lean out a bit while it is easy. That is not a call to chase every scale blip. It is a call to use a favorable phase while the data are aligned, then reevaluate when the phase changes.

This is where AI can help if it is disciplined, and hurt if it is eager. A decent coaching model should do three things before recommending a nutrition change:

Identify the time scale of the signal.
Match the change to the smallest variable that explains the signal.
Delay bigger edits until the signal survives a trend window.

Joe’s case shows why. One day of altered insulin response justified moving meals closer together and planning a dose reduction on the next high day. It did not justify rewriting the whole diet. Justin’s comment about being “at the half way point” of a planned six-week recomp also matters: he did not make a hard call from a single mid-point look. He acknowledged sickness, water retention, and the fact that fat loss can be hard to see in real time even when it is happening. That is the broader coaching principle: day-to-day visibility is poor, so the smaller the signal, the more time you need before acting.

The most common AI failure mode here is confidence inflation. The system sees a new pattern and treats it as a new rule. A client reports less appetite after one dose change, and the model wants to rewrite the entire food structure. A client sees a lower blood sugar after a high day, and the model wants to lower insulin across the board. But the evidence in these coaching exchanges points to a narrower practice: adjust the variable that moved, in the phase that makes sense, and give the rest of the plan time to declare itself.

For coaches, the useful question is not “Did anything change?” Something always changes. The useful question is “What changed fast enough to require a small edit, and what needs a full trend before I touch it?” That question protects you from two bad outcomes at once: under-reacting to a real timing issue, and over-reacting to transient noise. In a nutrition system, those are the same mistake in different outfits.

So the practical rule is straightforward. Make timing edits quickly when the variable itself is the problem: meal spacing, dose timing, food order, or how much food fits before appetite shuts down. Wait longer before changing the whole macro plan, the phase strategy, or the interpretation of progress. If a client’s sensitivity improves on a high day, move the meal. If appetite collapses on a small dose change, reconsider the dose. If the physique is in mid-recomp and the photos are messy, don’t confuse poor visibility with no progress. That is how you avoid over-adjustment: act fast on the lever that moved, and slow down everywhere else.

Sources Used

modules/03-knowledge/kahunas-coaching-deep-nutrition.md
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/joe_webb___members-rksigkykimaxwmo_t4_e8nwvbtc2j0etleutkyysads.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

Sources Used

/Users/justinharris/TroponinIQ/kb/supertrop/raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w13-18m/clients/rory_lazowski___members-c5balaovjbdoeefqmfuqdhh2tbpmfdu16lnf0tnrtmw.json
/Users/justinharris/TroponinIQ/kb/supertrop/raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w13-18m/transcripts/rory_lazowski___members-c5balaovjbdoeefqmfuqdhh2tbpmfdu16lnf0tnrtmw.md
/Users/justinharris/TroponinIQ/kb/supertrop/raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w13-18m/clients/joe_webb___members-rksigkykimaxwmo_t4_e8nwvbtc2j0etleutkyysads.json
/Users/justinharris/TroponinIQ/kb/supertrop/modules/03-knowledge/kahunas-coaching-deep-nutrition.md
/Users/justinharris/TroponinIQ/kb/supertrop/raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w13-18m/transcripts/joe_webb___members-rksigkykimaxwmo_t4_e8nwvbtc2j0etleutkyysads.md