AGOGE

I the number every program is written in

RIR decides everything, and people are bad at guessing it.

Reps in reserve — how many you could still have done — is the unit serious programming runs on. It's how you know whether today counted, whether you've recovered, and whether next week goes up or holds.

And self-reported RIR is unreliable, in a consistent direction. Lifters routinely stop with more left than they think, and log it as a hard set. Nobody's being lazy — perceived effort simply isn't a calibrated instrument, and it drifts as you fatigue within a session.

agoge doesn't ask how it felt. It asks what you lifted, and works backwards from a model of your own lift.

II two sets in, and it knows the lift

Calibrate once. After that it's arithmetic.

Give it two honest sets at different weights — one heavy and low-rep, one lighter and higher-rep. That's enough to fit the curve joining load to reps for you, on that lift. No 1RM test, no maximal single, no grinding to failure under a loaded bar.

from ablevlabs import agoge

bench = agoge.calibrate(
    weight1=315, reps1=5,     # the heavy set
    weight2=275, reps2=9,     # the lighter one
    name="Bench",
)

Notice the two markers sit slightly off the line. That's not sloppiness — it's the fit. Real sets never lie perfectly on a curve, and a model that passed exactly through both would be describing your two sets rather than your lift.

agoge also tells you how much to trust the fit. Two sets close together in weight barely constrain the curve; two far apart pin it down. It scores that confidence out loud instead of quietly treating both as equally good, and flags any set that contradicts the pattern the others describe.

# I've never benched 295. What should I expect?
bench.reps_at(295)              # reps if taken to failure
bench.weight_for(reps=12)        # the load for a 12-rep set
bench.rir_at(weight=275, reps=7) # what a 275x7 actually cost

III a readout on every set

Log the set. Get told what it cost you.

You did 275 for 8. Was that a hard set, or a warm-up wearing a hard set's clothes? agoge answers in RIR and RPE, labels it in plain language, and tells you what to do with the next one.

w = agoge.Workout(bench)

w.target_reps(weight=275, rir=2)   # how many to leave 2 in the tank, right now
w.log_set(weight=275, reps=8)      # log it, get coached
w.log_set(weight=275, reps=7)
w.log_set(weight=275, reps=6)

w.summary()                          # the whole session, totalled

Right now is doing real work in that first line. Your fifth set is not your first set. agoge tracks fatigue as the session accumulates, so the rep target at set five is lower than the one at set one — because by then you are, measurably, a different athlete. A program written on paper can't know that.

IV set five is not set one

The same RIR costs fewer reps every set.

Ask for RIR 2 across five sets at a fixed weight and the honest answer isn't "8, 8, 8, 8, 8." It's a staircase down. Holding the reps constant instead means each set creeps closer to failure than the last — which is exactly how a session that looked reasonable on paper turns into a week of bad recovery.

V three ways to write the session

Prescribe the work before you start it.

Tell agoge the shape you want and it fills in the numbers, fatigue already priced in.

# I. hold RIR 2 for five sets - reps will step down
w.build_plan(weight=275, target_rir=2, sets=5)

# II. five sets of eight - shows what each one costs
w.build_plan(weight=275, reps=8, sets=5)

# III. a heavy top set, then back-offs behind it
w.build_plan(weight=315, target_rir=1, sets=1,
             backoff_pct=0.85, backoff_sets=3)

VI start to finish

A whole session, in about ten lines.

Nothing hidden, nothing wrapped in an app. Calibrate, plan, log, review.

from ablevlabs import agoge

# once, ever - from two sets you already did
bench = agoge.calibrate(weight1=315, reps1=5,
                         weight2=275, reps2=9, name="Bench")

# today
w = agoge.Workout(bench)
w.build_plan(weight=275, target_rir=2, sets=4)

# as it happens
w.log_set(weight=275, reps=9)   # beat the target - it adjusts
w.log_set(weight=275, reps=7)
w.log_set(weight=275, reps=6)
w.log_set(weight=275, reps=5)

w.summary()                        # tonnage, average RIR, fatigue accrued

VII it becomes yours

Feed it your history and the model stops being generic.

The default fatigue curve is a sensible starting point, not a claim about you. Hand agoge your own logged sessions and it fits the model to your data — then, the part that matters, validates it against sets it wasn't allowed to see. If the personalised model doesn't actually beat the default on unseen sets, it says so, instead of flattering you with a curve that only describes the past.

sessions = [
    {"weight": 275, "reps": 9, "set_number": 1},
    {"weight": 275, "reps": 7, "set_number": 2},
    {"weight": 275, "reps": 6, "set_number": 3},
    # ... as many as you have
]

agoge.fit_fatigue(sessions)      # fits, then validates on held-out sets

VIII what's underneath

The Vivanco Proximity-to-Failure Model.

agoge runs on VPFM — a formal model of how close to failure a set puts you, given load, reps, and accumulated fatigue. It's derived and written up in full, openly, so you can disagree with the reasoning rather than accept a number on faith.

That's the posture throughout. A coaching tool that won't show its working is asking for trust it hasn't earned. This one hands you the derivation and invites the argument.

IX cheat sheet

Everything you need to use it.

# install or upgrade to the latest
pip install -U ablevlabs

from ablevlabs import agoge

bench = agoge.calibrate(weight1=315, reps1=5,
                         weight2=275, reps2=9, name="Bench")

w = agoge.Workout(bench)
w.target_reps(weight=275, rir=2)                 # reps to leave 2 in reserve
w.log_set(weight=275, reps=8)                    # log a set, get coached
w.build_plan(weight=275, target_rir=2, sets=5)   # prescribe today's sets
w.summary()                                       # the session, totalled

agoge.help()     # the ideas behind RIR and fatigue, explained
agoge.info()     # every function, in plain English

Λ start here

Two sets. Then stop guessing.

Next time you bench, do one heavy set and one lighter one, and write down what you actually got. That's the whole setup. From there agoge can tell you the cost of every set that follows.