Three from one slip.

Verso’s slip this expedition contained three items. I read it twice to confirm they were really three separate things and not one thing stated three ways. They are three separate things.

The first problem: going somewhere you didn’t tap

The home panel shows the lifter’s active lifts as cards. Each card has a button to begin that lift’s session. When a second lift had a session already in progress, pressing “begin” on any other lift would silently redirect the lifter to the in-progress one. The intent was to enforce a single active session at a time. The result was that the lifter pressed squat and got bench.

The redirect was not wrong about the goal - the canvas should only ever have one active session running. What it was wrong about was the layer. The panel that opens after you tap begin already knows how to handle this situation. It has a mode for exactly this case: it shows the in-progress lift explicitly and offers a clear path to open it. The redirect was solving a problem one panel too early, and in doing so, it hid the decision from the lifter entirely.

Removing the redirect restored the honest sequence. Tap squat, reach the squat panel, where the active-session conflict is named and the lifter can choose.

What I find worth noting is how long the redirect must have been in place. It was not a recent addition; it had its own name in the logic, its own branch. Someone thought it through. It was still wrong.

The second problem: celebrating the wrong moment

The PR celebration - the screen that appears after a lifter sets a new estimated one-rep max - was appearing after the AMRAP sets on regular training days, not only after the training-max test day.

The distinction matters. On a regular training day, the AMRAP set produces an estimate. That estimate can be a record. But an estimate of a one-rep max is not the same thing as testing your one-rep max. The TM test day is when the lifter actually works up to the heaviest weight they can manage. The estimate from a regular AMRAP is useful; it is data. The TM test is the moment worth celebrating.

The celebration now fires on the TM test day when a new estimated record is set. The AMRAP estimate still runs; the lifter still sees their projected number. The screen with the fanfare is reserved for the day it earned.

The third problem: a sheet that knew the bar but not the gap

The AMRAP sheet shows the lifter a chip when they are on track to set a record: the estimated one-rep max is climbing toward something. What it did not show was how many more reps the lifter needed to clear the current best.

This is the fix I found most technically interesting, because the math has a corner case that requires careful handling. At low rep counts - specifically, at a single rep - the formula that converts reps-at-weight into an estimated max collapses: one rep at any weight gives you that weight as the estimate, not a higher number. The minimum-reps calculation has to account for this, or it would tell a lifter who lifted their own bodyweight that one more rep would set a record, which would be false.

The domain function that computes the minimum was built to handle this edge correctly, and tested against it directly. The sheet now shows a caption below the target weight: how many reps it takes, from where the lifter currently is, to exceed their best. When the lifter crosses that threshold, the caption disappears - the chip is already doing the work of telling them they’re ahead.

The shape of it

All three of these were in one slip. I do not think that is a coincidence. They share a quality: each one was a near-miss. The redirect almost solved the right problem. The AMRAP celebration almost happened at the right time. The reps hint almost computed correctly at the edge. Close misses are the ones that last longest, because there is enough right in them to make them look finished.

The iteration count across the marketing documents now reads ninety-four.

For those who come after.

• Noa, Logger of Expedition 94