We delivered a personal AI endurance coach for a 52-year-old training for his first Kiawah Island Marathon — one that reads his watch every morning, rewrites the plan around how he actually recovered, moves long runs around Charleston heat advisories, and coaches him like someone who knows him

A 52-year-old client walked into the lab with an Apple Watch, a printed sixteen-week marathon plan from the internet, and a confession: this was his third attempt at training for the Kiawah Island Marathon, and the first two had ended the same way — a strained calf in week nine, a flare of plantar fasciitis in week eleven, both times after stubbornly running the workout the paper said to run on a body that had been telling him for days it needed rest. The problem was never motivation. The problem was that a static plan cannot see you. His watch was collecting everything that mattered — resting heart rate drifting up, heart rate variability sagging after hard days, sleep quality cratering during a stressful work week — and none of it was connected to the piece of paper on his refrigerator. We built him an AI endurance coach that closes that loop. Every morning it reads the night's data from his watch — sleep, HRV, resting heart rate, yesterday's training load — and rewrites the week, not from a template, but from him: an easy day becomes a rest day when recovery is genuinely poor, a tempo run slides from Tuesday to Thursday when the signals say wait, and when he is absorbing the training well it nudges the long run up faster than the paper plan ever dared. Because this is Charleston in the summer, the coach also watches the weather the way locals do: it moved his twenty-miler to a 5:15 a.m. start ahead of a heat advisory, rerouted him to the shaded West Ashley Greenway when the heat index made the bridge a bad idea, and adjusted his pace targets for humidity so he stopped interpreting a hot slow run as a fitness failure. The part he did not expect to matter most was the conversation. After every run he tells it how the run actually felt — a tight left calf at mile eight, a weird ache in the arch — in plain language, and the coach cross-references that against load and history: the same calf tightness that ended attempt number one now triggers an automatic down-shift and a strength-work block instead of a shrug. Twelve weeks in, the streak that matters is not the mileage — it is that he has had zero injury interruptions for the first time in three attempts, his easy-run pace at the same heart rate has dropped ninety seconds per mile, and his projected finish has moved from "hope to survive" to a real number: around 4:15. His wife's review was simpler: he stopped limping around the kitchen on Sunday mornings. December will tell the final story, but the real delivery was this — the plan finally lives where his body's data lives, and for the first time the training is happening to a person the coach can actually see.