Three Weeks of Coaching Memory

Three Weeks of Coaching Memory

AI fitness coaching scales fastest when it drafts, tracks, and flags—but judgment still has to decide

The 5% details compound into 5 lbs over a decade in Justin Harris’s coaching framework: if a client gains 10 lbs of muscle, a 5% edge is half a pound in a year, and that same edge becomes visible over time through repetition and adherence, not flashy interventions. That is the mechanism here: cumulative marginal gains. The falsifiable thesis is simple: AI coaching will create real leverage for rosters only when it handles the repetitive work that humans repeat a dozen times anyway, while leaving the coach’s judgment in charge of when to hold the line, when to adjust, and when not to overreact.

That thesis matters because the strongest evidence in the coaching KB is not “AI knows better.” It is that the highest-value coaching decisions are often small, repeated, and easy to standardize. In the Joe Webb transcript, a high day produced a noticeable drop in blood sugar on the same insulin dose that had worked the week before; the immediate fix was not a new philosophy, just a 1 iu reduction and a meal timing adjustment when the same pattern showed up again. In Rory Lazowski’s transcript, Justin explicitly says the hardest part of coaching is remembering repeated topics across clients and being hyper paranoid that clients will feel ignored, so he repeats things a dozen times just to be safe. Those are not edge cases. They are the daily mechanics of roster-scale coaching: notice the pattern, compare it to prior weeks, decide whether the change is meaningful, and communicate it in a way the client can actually use.

AI is well suited to the first and third steps. It can summarize prior check-ins, surface what changed since the last high day, and draft a clean reminder that says, in plain language, “same dose, lower glucose, shorten the interval next time.” That is leverage. A coach who manages 20 or 50 athletes does not need another generic dashboard; they need fewer missed comparisons and less time spent rebuilding the same context from scratch. The tool’s job is to make the coach faster at seeing what already happened.

But the transcripts also show where AI should stop. In Joe’s case, the signal was not just “blood sugar was lower.” It was that the day was already structured around a high day, the drop was noticeable but not a proper hypo, the client responded by moving meal 2 earlier, and then the same reduced dose still caused the same issue again on the third shot. The useful judgment is not a one-size-fits-all rule. It is deciding whether the change reflects improved sensitivity, a temporary fluctuation, or a need to wait and observe on the next high day before making another adjustment. That is judgment, not automation.

This is the real coach leverage equation:

1. AI compresses memory. The Rory exchange shows how much coaching bandwidth gets spent on repeated explanations and reassurance. A system that remembers the last discussion, the last change, and the last outcome removes a huge amount of duplication.

2. AI standardizes observation. The Joe exchange is basically a mini decision tree: same dose, lower reading, meal spacing changed, dose reduced, issue repeated. AI can structure that sequence so the coach does not have to reconstruct it from scattered chat logs.

3. AI drafts, but does not decide. The coach still has to know whether a trend is real, whether the timing matters, whether the client is sick, and whether a change should be held for one more week. The source material repeatedly shows coaching as pattern recognition under uncertainty, not formula-following.

That matters even more as rosters grow. Scale increases the cost of forgetting, but it also increases the cost of overcorrecting. A coach with a few clients can afford to rely on memory and intuition. A coach with a large roster cannot. Yet a large roster is exactly where bad automation becomes dangerous in a non-medical sense: it can make the coach sound confident before the evidence is there. If the system pushes every small fluctuation into a recommendation, it erodes judgment instead of amplifying it.

The better model is narrower and more disciplined. Let AI do the labor of continuity: track the repeated topics, highlight the last known state, compare current notes to prior notes, and flag when a pattern has repeated. Let the coach do the work that cannot be outsourced: determining whether the situation is a real trend, choosing how much to change, and deciding whether the client needs reassurance, restraint, or a new plan. In practice, that means the AI should be more like an excellent assistant and less like an oracle.

There is also a lesson here for how coaches should evaluate AI vendors. Ask whether the tool improves the quality of recall, the speed of comparison, and the clarity of communication. Ignore claims about replacing coaching intuition. The source material gives no support for that fantasy. The strongest coaching moments in the KB are not when someone invents a new framework; they are when a coach notices an expected issue, explains it clearly, and makes a small adjustment without drama.

That is why the most useful AI fitness coaching products will not be the ones that promise to “run the program.” They will be the ones that help a coach manage more athletes without becoming less attentive. Roster scale comes from memory and workflow; preserving judgment comes from keeping the final call human. If an AI system cannot do both, it is not leverage. It is just more noise.

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/joe_webb___members-rksigkykimaxwmo_t4_e8nwvbtc2j0etleutkyysads.md
• raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w19-24m/clients/rory_lazowski___members-c5balaovjbdoeefqmfuqdhh2tbpmfdu16lnf0tnrtmw.json
• raw/_consumed/2026-05-31/kahunas-export/2026-05-31-w19-24m/transcripts/rory_lazowski___members-c5balaovjbdoeefqmfuqdhh2tbpmfdu16lnf0tnrtmw.md
• modules/03-knowledge/kahunas-coaching-deep-nutrition.md
• raw/_TROPONIN_SENTIMENT/troponin_community_sentiment_kb.md