An AI coach for muscle gain: surplus, sets, protein and the log

The four levers, with the numbers

Lever	Target	Evidence
Energy	A surplus of roughly 10–20 percent; gain ~0.25–0.5 percent of bodyweight a week if you are a novice or intermediate, less if advanced	Iraki et al., 2019; Slater et al., 2019
Volume	About 10–20 hard sets per muscle group per week, with more volume producing more growth up to a point	Schoenfeld et al., 2017
Protein	About 1.6 g per kg of bodyweight per day, with little extra benefit above it; up to 2.2 g/kg is a reasonable ceiling	Morton et al., 2018; Iraki 2019
Progression	Add load or reps when the prescribed work is completed; most sets 1–3 reps shy of failure	Refalo et al., 2024

Energy

Slater and colleagues' review asks directly whether a surplus is required to maximise hypertrophy and concludes that while muscle can be built at maintenance, especially in beginners, a modest surplus helps and a large one mostly adds fat. Iraki's review of off-season bodybuilding nutrition puts a number on "modest": about 10–20 percent above maintenance, aiming for 0.25–0.5 percent of bodyweight gained per week for novice and intermediate lifters and less for advanced ones. For an 80 kg man that is 200–400 g a week on the scale, which is slow enough that you will not see it in a single weigh-in; you need the trend.

Volume

Schoenfeld's dose-response meta-analysis found that each additional weekly set was associated with more growth, with ten or more sets per muscle per week outperforming fewer. The practical range most coaches work in is 10–20 hard sets per muscle, higher for muscles you want to prioritise, lower when recovery is poor. Counting those sets is the single most useful thing a log can do for a gaining phase.

Protein

Morton's meta-analysis of 49 trials found protein supplementation increased gains in lean mass and strength during resistance training, with the benefit plateauing at around 1.6 g/kg/day. Iraki's recommendation is 1.6–2.2 g/kg, spread across meals. The protein guide works through what that looks like in food.

Progression

A set that does not challenge the muscle does not grow it, and a program that never adds load stalls. Refalo's review of proximity to failure finds similar hypertrophy when sets stop a few reps short of failure as when taken to it, which is the basis for training at 1–3 reps in reserve most of the time and adding weight by rule rather than by feel. Progressive overload, explained covers the rules.

What an AI coach does with those levers

Most muscle-gain content stops at the numbers above. The problem is not knowing them; it is that the four levers interact and drift. You eat more, the scale jumps from water and glycogen, you panic and cut. Your sets per muscle creep down as a block gets hard. Protein falls on busy days. A coach's job is to keep each lever in its lane and tell you, briefly, when one has left it.

Here is how Bay does it over text, free right now:

Energy. You set a gain goal and Bay sets calorie and protein targets. Every Sunday it compares your logged intake with your weight trend over the past 14 days and, if the observed rate of gain disagrees with the goal rate, proposes a change of at most 150 calories, protein unchanged, carbohydrates absorbing the difference. It will not propose anything until it has four weigh-ins and ten logged days, and you approve each change with a tapback. If your trend clearly contradicts the goal mid-week, it sends an off-cycle review rather than waiting. Ask "am I on track?" any day for the same analysis.

Volume. Bay's plans are built to volume bands by experience (8–14 hard sets per muscle a week for beginners, 10–18 for intermediates, 12–20 for advanced lifters), with the muscles you ask to emphasise at the top of the band. Text "show me my lifts" and the card that comes back shows your weekly hard sets per muscle against your band, so you can see a lagging muscle before the block ends.

Protein. Meals are logged by text, photo or voice note, with protein counted from label or menu data where it exists. The evening recap shows the day's protein against the target, and the single daily nudge, timed to when you usually eat, is the reminder that you have not logged. How that works is in how to track calories by texting.

Progression. Each training morning's text carries the session with a weight for each lift. Hit every rep and the next session's weight goes up by that lift's increment; miss twice and it drops ten percent and rebuilds. Plans run in five-week blocks with a lighter fifth week once you are past the beginner stage, and sets are prescribed at 1–3 reps in reserve. Log in shorthand ("bench 155x8x3") and the numbers for next time are set.

A sample gaining week

For an intermediate lifter, 75 kg, four days a week, with a gain goal:

Lever	What it looks like
Energy	Maintenance about 2,600 kcal; target 2,900 kcal (a ~12 percent surplus); weigh daily, judge the trend weekly; aim for about 0.3 percent (~225 g) a week
Volume	Upper/lower split, 14–16 hard sets per muscle a week; chest and back at the top of the band
Protein	120–150 g a day (1.6–2.0 g/kg) across three or four meals
Progression	Main lifts 6–10 reps at 1–3 RIR; +2.5 kg on upper and +5 kg on lower lifts when all reps are hit

If after four weeks the trend is flat, add 100–150 kcal. If it is above 0.5 percent a week, pull back by the same amount. The weekly review automates exactly that decision.