How it works
One loop. You ask, you train, you tell him how many reps you had left, he reacts.
Everything below is what the app actually does, written from the app's own code rather than from a brochure. Where it refuses to answer, that is a feature and it is described here too.
- On day one you and Jerry make a dealJerry asks your bodyweight, your experience, how many days a week you train, and your current lifts if you know them. Then he tells you plainly that on day one he has averages for athletes of your sex and training experience, adjusted for your age and bodyweight or based on the weights you gave him, and nothing he has measured himself. He asks one thing in return: after each set, tell him honestly how many more reps you could have done. Round up to look strong and he coaches a fake version of you.
- He builds the plan, then keeps score on himselfYou get a multi-week plan and a first session, labelled for what it is: true for the average athlete like you, not yet true about you. Every prediction he makes about you is later checked against what you actually did, hits and misses alike.
- Three ways inLet him plan the whole session, pick the muscles you want, or freestyle and log whatever you do. If you pick muscles he does not simply obey: if a muscle still needs recovery, or your weekly workload is already high, he trims or declines it and shows you a short note explaining exactly why.
- Tell him how many reps you had left, and he reacts before the next setAfter every set he compares your answer against what he expected and tells you in one line whether you are on track, drifting, or still showing him what you can do on a new lift. If you are fading faster than planned he cuts the remaining sets on the spot and says that the cut protects your recovery.
- The weight goes up when you have earned itNot on a calendar. He holds the load and asks for one more rep, then adds the plate once the reps are there. Outside the main lifts of a program built for strength, or for strength and size together, finish every set he asks for two sessions running and he stops waiting: he takes a bigger step, never more than a tenth of the weight in a week, and goes back to small steps at the first set you cannot finish. It runs on the reps you actually completed. When you type in how many more reps you had left, that can hold the weight back, and it can push the weight up only if you typed that number yourself. Tapping to accept the number he suggested never pushes it up.
- Your guess of reps left does not need to be perfectJust make it honest. For each lift, Jerry compares your guesses with how your sets actually go. Once he has enough of your sets, he corrects the weight for the way you guess. If you keep matching your reps on a weight you called your limit, he reads that as reps left in the tank and adds weight.
- Tell him something and see what changesReport an injury and he tells you how many of your movements are now held out or lightened, starting with your next session. Say a muscle is sore and he shows you how much easier he made your next sets. Travel, get sick, change your equipment, and the plan changes, with a note saying why. If you quietly use a lighter weight than he asked for, he notices, and if you do it again on another day he lowers the weight he gives you himself instead of arguing with you.
- It gets more personal over timeEvery number starts as an honest average. As you log real sets he moves from averages to your own numbers and shows you how far he has moved and how sure he is. How fast that happens depends on what you log: sessions with a lot of real working sets move him onto your numbers faster than the calendar does. The home screen switches from Building to Individual only once you have at least 30 workouts, spread across at least eight different weeks, with five lifts trained on eight or more days each, and you have trained through two phases. Three of the deepest analyses are the exception and wait for a fixed minimum of fifteen sessions' worth of data: fifteen workouts, or enough good working sets to equal them, where one long day counts as two at most and one week as ten at most. Below that there is too little data to mean anything. A fourth, the story of how you have changed, also waits 60 days from your first set. The app names each one that is still waiting on the screen where it would appear.
Four moments
Four things Jerry does that a logbook cannot.
Hard Truths
When your own data shows a habit working against you, he says it out loud, using only your logged numbers. Before he speaks he rules out the innocent explanations: an injury, a beginner still learning, a planned easy week, a comeback. Each time he speaks up, it comes with a note saying whose data it rests on, how much of it there is, and what he checked first. If he cannot rule out the innocent explanation he stays silent, because one false accusation costs trust permanently.
The Principle
Instead of a generic tips feed, he teaches one principle at the exact moment your own training demonstrates it, and cites the research behind it. One thing at a time, never repeated, never invented. The lesson itself carries no numbers, because your numbers are in the note directly above it.
The day you get hurt
This is where other apps go quiet. Tell him something hurts and he keeps coaching: removes what is unsafe, trains everything else, and brings the injured movement back a step at a time. Pain above 5 out of 10, the line rehabilitation research uses, holds it at its current step rather than letting the calendar advance it. The screen shows the honest number of days until full load, and it is the same number the app is actually working from. This is training around an injury, not treating it. It is not rehabilitation. A qualified healthcare professional decides what the injury itself needs.
The set he read
Wear your Apple Watch, your AirPods or both, and he reads your sets with nothing to tap. The watch feels your hands move and the AirPods feel your head move, so between them they read presses, rows, curls, squats, deadlifts, pull-ups, dips and push-ups. He counts your reps and measures how much you slowed down from the first rep to the last, the clearest sign of how close you came to failure. Until he has checked each device against what you log, lift by lift, he shows you what it read and draws no conclusions. Until a device has proved itself on a lift, nothing it reads changes the weight on your bar. After that, if what it measured on your first set of that lift has predicted your day better than your sleep and heart-rate variability did, it can lighten that lift, never by more than 15 percent. The heart rate your watch reads during a set never moves the bar.
On the screen
The number on the screen is the number the app is actually using.
When you report pain, your return to full load does not move ahead on a calendar. Pain above 5 out of 10 holds it where it is for a week, and another report holds it again. There is no separate optimistic figure shown to keep you feeling good, because a comforting number you cannot act on is worse than a hard one you can.
Every hard truth comes with a note in the same style: whose data it uses, how much of it, and what was ruled out before he said anything. In the examples, population figures are averages taken from many athletes rather than from you, and a deload is a planned lighter week.
Knee
Today: 40% of full load on this area's main lifts, and full load in 17 days. A pain report above the line pauses the rebuild for 7 days, and it resumes on the date Jerry shows you. Another report above the line holds it again.
The work you route around
Quadriceps, Hamstrings: under the growth minimum for three weeks while Chest sat at full volume.
From your own logged weeks, the last three, weighted by how much each lift involves the muscle. The minimums are population figures.
Checked first: your injury log (clear on the lower body), no declared specialization, not a deload or a return from time away.
Example, not a screenshot. The sentences are the app's own, except the date.
The part that matters most
He would rather refuse than guess.
Most coaching apps promise more than they can back up, and once you catch one of those promises you quietly discount everything else. This one does the opposite: it marks every guess as a guess and counts every miss, so the confident claims actually carry weight.
Saying nothing is built in more carefully than speaking is. When he cannot see enough to give you a number, he says so and names what is missing, instead of guessing. Your watch and AirPods draw no conclusions until he has checked them against what you log, and if the two read the same set differently, he leaves it unmeasured instead of picking one. Ask him a question in plain language and every figure in the answer has to trace back to a number the app really calculated: if one does not, the whole answer is thrown away and a plainer one takes its place.
And one part of the app is barred from saying anything about you at all. It is a recovery estimate that can only ever be checked against numbers it made up itself, so it is labelled as exactly that and is kept out of every claim about you. It can never be proven wrong, which is why it cannot count as evidence.
Your data
Easy to leave, hard to replace.
Your training is computed on your device by the app's own code. Bring your history in from the big logging apps, so switching does not mean starting at zero. Export your whole history, back to your first set, as one file whenever you want. The export is free. Nothing is held hostage to keep you subscribed.
Losing your history is one of the loudest complaints in this category, and making cancellation difficult is what gets an app called a scam. We would rather be the product that is easy to leave and hard to replace.
Exactly what does and does not leave your phone, including the seven things that can leave it and the one of those we have removed, is set out in the privacy policy.
Try it yourself
Want it on your phone this week?
It is in testing, and testers get the whole app free. Send one message and a person writes back.
You do not have to trust any of this page. Try to break a sealed prediction, or read the record.