AI Powerlifting Coach: RPE, Peaking, and Meet Day

Powerlifting is three lifts and one day. Everything between now and that day is an argument about load and fatigue, and most of it gets settled by numbers you already produce every session and then throw away. ChatRPE is an AI powerlifting coach in the narrow, useful sense. It takes the set you just did, keeps the weight, the reps, and the effort rating together, and answers "is this working?" from your own logbook instead of from a program's promises.

What follows is the actual mechanism: the shape your data takes, the charts that come out of it, and the decisions those charts settle. If you want the general strength argument first, how to get stronger has the levers and the rep ranges, and the RPE guide covers what an 8 means in reps left in the tank. Send a photo or a clip of a set and you also get a read on what is visible in the frame, which the form check guide covers.

Every set carries its RPE, and that is the foundation

Most lifting apps treat RPE as an optional note field. In ChatRPE it is part of the structure of a strength entry. The saved shape is { date, exercises: [{ name, sets: [{ weight, reps, rpe }] }], notes }, so effort is a first-class number sitting next to load and reps on every set.

Everything below depends on that one detail. A rating buried in free text can be read back by a human. A rating stored as a field can be averaged, trended, filtered by lift, and charted across a block.

Say it out loud rather than typing between sets. On iOS you can send it without unlocking the phone, which the hands-free logging guide covers. Notes are keyword-searchable across your whole history, so "when did my elbow last flare up" has an answer.

Try this: "Log squat 210×3 at RPE 8, 215×3 at RPE 9, then tell me my average top-set RPE per week for the last four weeks and whether it's creeping up."

Your estimated max, derived at draw time from sets you already did

The number a powerlifter wants is the one they have not tested. ChatRPE computes it rather than asking you to store it, and the mechanism is worth knowing, because it is what separates this from a tracker with a 1RM text box.

A chart can evaluate a formula over the fields of each data point as it draws. Point it at one named lift with a path filter, apply the Epley estimate, and take the highest value per session:

• path: exercises[?(@.name=="Squat")].sets[*]
• formula: weight * (1 + reps / 30)
• aggregation: max

That is one request, not a workflow. It also runs retroactively over sets you logged months before you thought about charting them, because nothing had to be pre-computed. The practical consequence is that you never have to max out to know where you are. Twelve weeks of triples and fives produce an estimated max curve, and the shape of that curve is the thing to program against.

Two things stack on top of it. A trend line overlay shows direction instead of the sawtooth. A target line at a specific value with a label puts your goal third attempt on the chart as a horizontal reference, so every session's estimate is read against the number that actually matters.

Estimates degrade as reps climb. A five is a decent input, a ten is a poor one, and a grinder at RPE 10 flatters itself. Compare this month against last month at similar rep ranges rather than treating any single value as a max you could hit today.

Try this: "Chart my estimated 1RM per session for squat, bench, and deadlift over the last six months with a trend line, and put a target line on the squat at 250 kg."

Weight-class descent, with the number that matters on your sidebar

Making weight goes wrong the same way every time. Nobody watches the trend until eight days out. Two things help.

The first is reading bodyweight as a moving average, the noisy daily number smoothed over a window, with a target line at your class limit. The chart then answers "am I going to make weight" as a slope against a line, not as a number you have to interpret. Charts of irregularly dated weigh-ins can be forced to proportional time spacing, so a gap of nine days looks like a gap of nine days.

The second is a persistent dashboard. A KPI widget takes a data type, an aggregation, and an optional target with a remaining operation, so "2.6 kg to go" sits in your sidebar as a live number next to a chart of the descent and a list of your recent top sets. On an iPhone that tile can also go on your home screen. You stop having to ask.

Put the descent on the same chart as your estimated max and you get the finding that changes decisions. A stalled estimate against a falling bodyweight is the moment to stop the cut, not push it. The weight loss tracker guide covers reading a bodyweight trend properly.

Aggressive cuts close to a meet cost strength, and everyone believes they are the exception. If you are cutting hard, sitting near a class boundary, or under eighteen, make the plan with a coach and a clinician, then use the descent chart to hold yourself to it.

Try this: "Chart my bodyweight for the last 90 days with a 7-day moving average and a target line at 83 kg, then add a KPI to my dashboard showing how far I have left to cut."

The meet date as durable context, not a thing you keep repeating

Tell ChatRPE the date and it can hold it as a stored goal or constraint: a small, user-approved, expiring piece of context that proactive coaching reads later instead of losing when the conversation ends. "Meet on November 8th, 83 kg class, opener target 220" survives into next month's sessions.

That is also what makes reminders useful rather than annoying. A recurring reminder can carry a condition evaluated against your own data before it fires, so a Thursday prompt that only arrives when you have logged fewer than two sessions that week suppresses itself in the weeks you were already on track. The accountability guide covers the mechanism, including that delivery is push through the iOS app.

Peaking, and the week nobody logs honestly

The last four weeks are where the logbook stops being a diary and becomes the plan. Volume comes down, intensity goes up, and the whole thing depends on an accurate read of fatigue, which is the input people fudge when they are excited.

Three things make that stretch readable. Average session RPE as its own line is the first: effort rising while load stays flat is the classic sign of accumulating fatigue faster than you adapt, and it is a chart rather than a feeling. Second, sleep and top-set RPE on one picture, two metrics on two axes in a single request, explains more bad peak weeks than programming does. The sleep tracker guide covers logging that side.

Third, run the heavy singles as a live session. You get an ordered list of openers with a rest countdown that restarts as you advance and a display line the coach rewrites each set, so the sequence sits on the screen instead of in your head between attempts. The guided workout app guide covers the flow.

A taper is not a deload, but the same logic applies. Planned reduction beats reactive collapse, and deload week covers making that call.

Picking attempts from data instead of nerves

Attempt selection is where a training log pays for itself. The conventional shape, an opener you could triple and a third that is a real attempt, only helps if "could triple" is grounded in something.

It is, if you logged the year. Ask for a summary instead of a picture and you get the average, peak, and direction of your estimated max over a window you name, computed from the same formula as the chart so the numbers agree. Add what you have actually done at RPE 8 and 9 in the last eight weeks and three numbers with reasons attached is a reasonable thing to ask for. You still make the final call in the warm-up room, where the information is your body rather than your phone.

Try this: "My meet is in three weeks. Summarise my estimated squat max trend since January, list every squat set I've logged at RPE 9 or above in the last eight weeks, then suggest an opener, second, and third, and tell me what would make you lower each one."

Log the meet itself when it is over. Nine attempts, made or missed, with the RPE you would give each one. That entry is the most valuable single record in a powerlifting logbook, and the one most likely to exist only as a memory and a phone video. Twelve months later it is the reference point for the next peak.