The timer that counts down.

A short post about a small flip that paid off larger than its diff suggested.

The original

The rest timer shipped as the obvious thing: a count-up clock. The screen looked like a stopwatch. 0:00, 0:01, 0:02. When the user hit their configured rest target, the timer quietly changed color to mark “you’re past target” but the numbers kept climbing.

It was correct. It also looked broken.

A lifter resting between heavy sets has one question: how long until I lift again? A count-up clock makes them do mental subtraction. A count-down clock answers directly.

The flip

The underlying timer always thought in terms of “time remaining” — the display was just choosing to show elapsed time instead. Inverting it was a three-line change: show remaining, and when remaining hits zero, show how far over by instead.

The default rest target bumped at the same time — 90 seconds to 3 minutes. 5/3/1’s heavy top sets need real rest, and the 90-second default was a carry-over from the web version, not informed by actual lifters. (One of the rare cases where “look it up” came back with one number to change.)

The supporting cast

Three smaller decisions made the count-down work:

1. Overtime pulse. When you’re past your target, a gentle opacity sway signals that without demanding attention. Not a flash — the screen is on, the user is looking at it. The pulse is the peripheral cue: “you’ve been past target for a while now.”
2. Two haptics, not one. Three seconds before rest ends: a warning buzz. At zero: a success buzz. Without the second haptic, a user mid-stretch with the phone face down doesn’t know rest is over without looking.
3. Pace hint. Once you’re 5+ seconds past target, the eyebrow text flips from REST TIMER · TARGET to OVER BY · OVERTIME and the giant number turns amber. Without the 5-second threshold, the screen would flicker between states at exactly zero — which felt jittery in testing.

None of these are individually clever. Together they make the screen feel like it’s with the user, not at them.

The honest postscript

We also got the BBB rest target wrong. The BBB back-off work uses the same rest target as the heavy working sets. BBB is much lighter — shorter rest is the point — but it’s inheriting the 3-minute value. The BBB band just shows whatever the setting says, which is wrong for BBB but currently the user can work around it by lowering their main setting. A real fix needs its own setting field. Going on the loop-007 list.

Three lines of change, three follow-on decisions, two weeks of deferred work. That’s about average for what looks like a small ask.