NEW: Ready for iPhone Duo
- Folded, the map controls move into the side column where iPhone Duo keeps its buttons, so the chips and dashboard get the full width of the outer screen
- Opened in landscape, the ride and replay screens use the side-card layout instead of the portrait dashboard stretched across the map — data card at the top, Start/Pause at the bottom, map between them
- On the open Duo, slope and lean sit directly under the data card, with the puck status and any warnings stacked beneath them — one column down the left, so the map stays clear
- The compass sits above the map controls instead of behind the dashboard — or on the clock, on the outer screen

IMPROVED: Missing heart rate now says why
- When the watch workout fails to start, the ride screen says "Watch workout failed — open Tripper on watch" instead of the generic "check watch" — the two have different fixes
- The watch now reports its workout's lifecycle (started, failed, first reading) back to the phone, so a ride that comes home without heart rate finally leaves a diagnosable trace
- This release makes the failure visible rather than claiming to cure it: the gaps come from the watch's own sensor under-sampling, and the trace is what the cure will be built on

FIXED: The slope reading flickered between steep up and steep down while stopped on a hill
- Stopped on a 30-degree climb, the card alternated between about +65 and −54 per cent roughly once a second — 1115 swings between up and down in a single ride, and GPS Sync burned the flicker into the exported video
- Two unrelated things looked identical to the app. Slope deliberately withdraws a single reading when something jolts the bike, and a rider shifting their weight at a stop does exactly that. Every withdrawal handed the display back to the puck's motion chip — the very thing the app's own slope was built to replace, and upside-down on top of it (below)
- A withdrawal now falls back to the reading from a moment ago, up to four seconds old. It is a refusal to answer for this instant, not an absence of any answer, and the honest stand-in is what the same instrument said a second earlier
- The chip is used only when the app has never had a reading at all — a first-generation puck, or the first second of a fresh link

FIXED: The puck chip's own slope was upside-down
- Nose-up read as nose-down. The maths that turns the chip's reported orientation into an angle assumes one axis convention; the puck uses another, so pitch came out inverted
- Roll came through correctly, which is exactly why this hid for so long: half the attitude was right, and the wrong half was the half the app had already stopped using
- Measured against gravity over 4728 stopped rows — a still bike makes the accelerometer a true inclinometer, because there is no acceleration to confuse with tilt — the chip's pitch tracked the real angle at −0.98 times it. A clean sign flip, not a drift or a bias
- The app's own slope replaced this channel several versions ago, so it only ever showed through as the fallback above. It was also inverted in the pitch column of exported ride files, which is now correct

FIXED: A spinning wheel could throw the grade to 174 per cent
- Slope subtracts the bike's own acceleration before it reads gravity, because braking pushes you forward and an accelerometer cannot tell that from the bike tipping nose-down. That acceleration was taken from wheel speed
- On a loose climb the wheel spins: it reports 0 to 40 km/h and back while the bike has not moved. The app read that collapse as a full g of braking that never happened and drove the grade to 174 per cent — on a pitch the accelerometer itself was holding steady at 33 degrees
- The correction now applies only while the bike is genuinely rolling, and can never be larger than the jolt the accelerometer actually felt over the same instant. Real braking shows up there; a wheel spinning under a stationary bike does not
- Scored across four recorded rides against gravity at standstill: average error 1.32 down to 0.56 degrees, worst case 28.3 down to 11.6, and readings more than 5 degrees adrift fell from 6.6 per cent of rows to 1.3
- Removing the correction altogether scores better on the average and worse where it counts — one stop needs it and goes from 0.2 degrees of error to 21 without it — so it was fixed rather than dropped

FIXED: Slope ramped smoothly up from flat after a wheelspin spike
- Slope has two sources and blends across the handover so the number does not jump. It blended from zero whenever it had no current reading, which means it climbed smoothly out of an imaginary flat road: one spike bounced the source twice in three seconds and the reading ramped 0 to 33 degrees each time, looking for all the world like a hill arriving
- It now blends from the last real reading. Where there is no real reading to blend from, it steps straight to the measured one — a step onto something measured beats a smooth climb out of a number nobody measured