We don't guess.
We measure.
TreadPilot drives a real treadmill belt, so its behaviour is not a matter of taste. Every decision is backed by research — published science, what commercial treadmills actually do, and simulations calibrated against real workout logs — and every note here is the write-up: in plain language, charts included.
A plan built around you
How running history, a measured profile, your goal and your available time become a varied training plan. Inside the calculations, the treadmill execution and the feedback loops that keep one tired evening from becoming a new fitness level.
Read it →The first live run: the research meets the belt
Two Lab notes made measurable promises about heart-rate control. A 52-minute run on real hardware measured both: zone entry in 5:59, right on the simulated six-to-seven-minute promise, zero overshoot, 94.5% of the time in band with a single correction — and one lost command that repaired itself in a second.
Read it →Before the governor: why heart-rate control needs a warm-up
A cold heart-rate reading is a delayed answer, not permission to hurry the belt. The planned method uses a visible 6–10 minute load ramp first, keeps every reducing safety response live, and only then lets fresh heart-rate data earn upward control.
Read it →Speed or slope: which dial should follow your heart?
Keeping pace fixed while the hill moves sounds smoother. Under the explicit assumption that one level means one percentage point, it may be a much coarser load step than a 0.2 km/h speed nudge — so slope must first earn its place on real hardware.
Read it →Reaching the zone should not take 20 minutes
A dropped Watch signal sent one real workout to a 3.0 km/h crawl, and the controller took 20 minutes to reach the 130–145 zone. The log shows why, and three planned changes cut that to about six minutes in simulation — with zero overshoots in 210 stress runs.
Read it →Inside the zone: where the belt will park your heart rate
A heart-rate segment rode the bottom edge of its band for a whole workout — the rule working exactly as written, and written wrong. The next update will push to the lower-middle of the band and then go quiet: a better parking spot which, in simulation, costs no extra speed changes.
Read it →Numbers from real workouts are labelled as measured; numbers from our calibrated simulations are labelled as simulated. Worked examples use explicitly illustrative data. Every research track is adversarially fact-checked before it lands here — and when a check corrects us, the corrected number is the one you read.