Repsy

What weight to start with

Nobody knows their load on day one of a new exercise. The mistake is not starting light — it is staying light for weeks because the search moves in fixed 2.5 kg steps.

Why "just add a little each time" is expensive

Adding a fixed amount per session treats week one as progression. It is not: at the start you are SEARCHING for the load, not advancing it. Searching by always adding the same step is the slowest search there is — the further you started from your real load, the more sessions you burn before the first real stimulus.

The search multiplies, and slows down

If the set felt easy, the next one does not get a fixed step: it multiplies. The jumps are ×1.6, ×1.4, ×1.25 — large at first, because uncertainty is large, and smaller as the load closes in. Same idea as guessing a number between 1 and 100 by halving the range instead of counting up by one.

The ramp, set by set

Every row is generated by the same function that suggests the weight inside a workout in the app. The scenario is "easy on every set" — the longest path the ramp can take.

Started atSet 1Set 2Set 3Set 4
10 kg10 kg16 kg22.5 kg28 kg
15 kg15 kg24 kg33.5 kg42 kg
20 kg20 kg32 kg45 kg56.5 kg
30 kg30 kg48 kg67 kg84 kg
40 kg40 kg64 kg89.5 kg112 kg

Read the 15 kg row: the session ends at 42 kg. At 2.5 kg per session, reaching the same point would take 11 sessions — weeks of training below what the body could handle.

Silence is not "easy"

The ramp only jumps when you say the set felt easy. A set with no answer locks the load where it is — it does not climb on the benefit of the doubt. That is deliberate: it makes "not sure" a safe button. An app that reads silence as permission to go up turns the first session into an injury risk, and erring high costs far more than sitting one step low.

What counts as easy

Effort is logged on a 6-to-10 scale, and only 6 counts as "felt easy". From 7 up you are already in the zone where the stimulus happens — jumping from there would reopen a search that had already finished.

The search ends the same day

There are at most 3 jumps per session. After that the ramp locks, even if it still feels light: a whole session of attempts is not a workout. And the load you found is not lost — next time, that set opens exactly where the search stopped.

FAQ

Is that safe? A 60% jump sounds like a lot.

The big jump only follows a load you just lifted and called easy, and only in the session where the load is still unknown. Multiplying 10 kg by 1.6 gives 16 kg — in absolute terms a small jump, taken exactly when uncertainty is highest. As the load climbs, the multiplier drops.

What if I cannot judge the effort?

Not answering is a valid answer: the load locks. You repeat the same weight on the next set and decide later. The ramp never climbs on its own.

Do I need extra test sets?

No. The search uses the sets the workout already called for — the first set is the attempt, and its answer sets the second. No extra sets, no invented warm-up.

Does it work for barbell lifts?

It does. The empty bar weight enters the math before the search starts, and what the ramp returns is converted into plates per side. If the bar has no known weight — the Smith machine case — the app asks once instead of guessing.

Does it start over on the second session?

No. There is no stored calibration state because none is needed: next time, each set opens at the weight it logged last time. The search happens once per exercise, and history takes over from there.

The first session already counts as training

Repsy builds the plan, suggests the weight for each set, and adjusts inside the session when you say it felt easy. You can start in the browser, with nothing to install.

Keep reading