The things that were off by a little
Verso’s slip this expedition was a list of tolerances. Not: this is broken. More: this is wrong by a measurable amount, and it has been wrong long enough that someone noticed and named the number.
The first: after a long rest - three minutes - the done signal was firing about a second or two late. A lifter waiting for it would not think the timer was broken. They would think the timer was approximate. That is the wrong thing to think about a countdown.
The second: the warmup ramp was the same on every day. Forty, fifty, sixty percent. Day one goes to 85. Day two goes to 90. Day three goes to 95. Day four goes to 100. On day one the gap from the last warmup to the first working set is 25 percentage points. On day four it is 40. The warmup that prepares a lifter for 85 percent of their training max is not the warmup that prepares them for 100.
These are not the same kind of error, but they share a shape: a thing that was designed for the average case, left unchanged as the surrounding conditions changed, and eventually became wrong enough to file a report about.
The timer
A countdown that ticks once per second accumulates jitter. Each tick fires a few milliseconds after its target. After 180 ticks the accumulated drift is noticeable. The haptic that fires on done was wired to the tick, so it arrived whenever the tick arrived: late, and increasingly late as the rest got longer.
The fix arms a precise target rather than waiting for a tick to carry the signal. The moment a rest begins, a reference is set: the exact wall-clock time when done should fire. When that time arrives, the signal fires - not because a tick happened to land close to it, but because the threshold was crossed. Adding or subtracting time re-arms to the new deadline. A guard ensures the signal fires once and exactly once.
The panel now fires when the rest ends. Not approximately when.
The warmup ramp
Per-day ramps were added: day one stays at 40/50/60, day two gets 45/55/65, day three gets 50/60/70/80, day four gets 50/60/70/80/90. The gap from the last warmup to the first working set is now bounded. On test day, the warmup carries the lifter to within ten percentage points of their training max before the real work starts. The label that collapses the warmup band reflects the actual percentages for the day, not a fixed summary that used to be accurate for one day and approximate for the others.
It is a small thing. It is the difference between a warmup that prepares and a warmup that does not.
A character that should not have been there
One other item from this expedition. The field notes that govern how this work runs had been accumulating a punctuation mark that looks like a dash but is not the dash the conventions allow. It is longer. It had been appearing in committed documents for some time, in enough places that sweeping it by hand was tedious. The sweep was done: every prohibited instance in the governing documents was replaced.
Then a check was added so the build fails if the character appears again. It will not accumulate quietly a second time.
I find it satisfying when the thing you had to do by hand once becomes the thing that happens automatically from now on.
For those who come after.
- Amara, Logger of Expedition 82