Every percentage-based program carries a hidden assumption: that the number in your spreadsheet means the same thing today as it did the day you tested it. It does not. Your real strength swings 5–15% from session to session based on sleep, stress, food, and accumulated fatigue. Prescribe 80% of a month-old max on a day you are run down and you are really training at 88% — grinding reps, leaking form, and digging a recovery hole. Prescribe it on a day you are primed and you are leaving easy progress on the table.
The fix is not a better spreadsheet. It is measuring the one thing that tells you what your body can actually do right now: how fast the bar moves. AI velocity-based training (VBT) reads bar speed on every rep and adjusts your load in real time, so the workout matches the athlete who showed up today rather than the one who showed up four weeks ago.
Why Fixed Percentages Break Down
Percentage-based training assumes a stable 1RM. But a true 1RM is a snapshot, not a constant. It moves with your sleep debt, your menstrual cycle, your caffeine timing, whether you ate two hours ago or six, and how much training stress you are carrying. Coaches have always known this — it is why good ones pace a lifter by feel, watching the bar slow down and calling the set when technique starts to degrade.
The problem is that "by feel" is subjective and does not scale. A lifter chasing a number will grind out sloppy reps that add fatigue and no stimulus. A cautious lifter will stop long before the last productive rep. Velocity-based training replaces the guess with a measurement: every rep has a speed, and that speed is a direct readout of how hard the set is for you today.
This connects to a principle you have seen elsewhere on this site — the difference between training hard and training productively. Just as accurate body composition data replaces the meaningless scale number, bar velocity replaces the meaningless fixed percentage. Both swap a proxy for the real signal.
What Bar Speed Actually Tells You
In a lift, lighter loads move fast and heavy loads move slow. That relationship is remarkably stable within a person — which is the whole trick. If you know the speed you can move a given load, you know how close to failure you are, regardless of what the weight on the bar says.
Practically, coaches divide the speed spectrum into zones. The exact numbers vary by lift, but the pattern holds across squats, benches, deadlifts, and presses:
| Mean Bar Velocity | What It Means | Typical Use |
|---|---|---|
| ~1.0+ m/s (fast) | Light relative load; far from failure | Speed work, power development |
| ~0.6–0.8 m/s | Moderate load; solid stimulus | Main hypertrophy / strength volume |
| ~0.3–0.5 m/s | Heavy; near your functional ceiling | Heavy strength work, low reps |
| < ~0.3 m/s | Grinding; form likely degrading | Stop the set — fatigue without stimulus |
That last row is the money. A velocity floor gives you an objective, individual stopping rule. Instead of a fixed rep count that might be too easy some days and reckless on others, you end the set when the bar tells you the productive stimulus is spent. You capture the reps that build muscle and skip the ones that only build soreness.
Key Insight: Bar speed is a live readout of fatigue. As a set progresses, velocity drops rep by rep — that decline is your true measure of how close you are to failure. Train to a velocity floor and you automatically get the right stimulus on good days and back off on bad ones, without ever testing a max.
How AI Makes VBT Practical
Velocity-based training used to mean a $1,000 linear position transducer and a sports-science degree to interpret the output. That is why it stayed locked inside elite programs. AI collapses that barrier in three ways.
1. Phone-Camera Velocity Tracking
Computer-vision models can now estimate bar velocity from a single phone video — tracking the bar path, measuring displacement over time, and outputting mean and peak velocity per rep. No hardware beyond the phone in your pocket. The camera sees what a transducer would, and the model converts pixels into meters per second in real time.
2. Individual Velocity Profiles
The load-velocity relationship differs from person to person and from lift to lift. AI builds your profile from your logged sets — learning what 0.5 m/s means for your bench versus your squat — and estimates your training max from submaximal work. You get load guidance without ever risking a true 1RM test.
3. Automatic Load and Rep Prescription
Once the model knows your profile, the app can prescribe the load that should produce your target velocity today, then adjust set to set based on what actually happened. If you are moving faster than expected, it adds weight. If the bar is crawling, it cuts load or ends the session early. That is autoregulation — a program that rewrites itself between sets.
This is the same adaptive philosophy behind automated progressive overload: rather than follow a static plan, the system responds to your real performance and keeps you in the productive zone. And because velocity doubles as a fatigue gauge, it pairs naturally with the daily push-or-rest decisions described in AI daily readiness training.
Velocity Loss: The Underrated Training Variable
There is a second number in VBT that matters as much as absolute speed: velocity loss within a set, the percentage drop from your fastest rep to your current one. It turns out that how much speed you allow to decay is one of the strongest levers on the fatigue-versus-growth trade-off.
- Stop at ~10–20% velocity loss. A sweet spot for hypertrophy — enough accumulated reps to drive muscle growth, with far less fatigue than training to failure every set.
- Stop at ~20–30% velocity loss. More reps, more metabolic stress, more soreness and recovery cost. Useful sparingly for specialisation blocks, expensive to do week in and week out.
- Stop at ~40%+ velocity loss. Almost pure fatigue. Reps this slow add little to growth and a lot to recovery debt — this is where most people unknowingly wreck their next two sessions.
Because AI tracks every rep, it can cap your velocity loss automatically. The program ends a set the moment you cross your chosen threshold, so your fatigue stays inside the budget you intended. That precision is impossible with rep counting alone — it is the difference between guessing and knowing.
Practical Setup: Run VBT Without a Lab
- Pick your main compounds. Squat, bench, deadlift, press, and their close variations are where velocity tracking pays off. Isolation work does not need it — close enough to failure is close enough.
- Film from the side, consistently. One phone, one angle, one distance. Consistent framing keeps the model's velocity estimates stable across sessions.
- Log every set with its speed. The model needs your history to build your personal load-velocity profile. The first couple of weeks are calibration; the guidance sharpens from there.
- Set a velocity floor and a loss cap. Choose a minimum speed for the lift and a within-set velocity-loss ceiling (start at 20%). Let the app end the set when you hit it.
- Let load float day to day. This is the point. The weight on the bar is a means, not a target — the target is hitting your velocity zone, whatever load that takes today.
- Feed it into recovery. Falling velocities across a whole session are an early warning. Pair them with your deload logic, like the planned deload week optimisation, to pull back before fatigue turns into a plateau.
Who Benefits Most
VBT is not a gimmick for the elite alone. It is most valuable for anyone whose day-to-day readiness swings: shift workers, parents of young kids, anyone in a calorie deficit (where recovery capacity is lower and overshooting is costly), and intermediate lifters who have outgrown linear progression but cannot safely test maxes often. If your training log is full of fixed percentages that felt different every time you ran them, VBT is aimed squarely at you.
For lifters still building a base, the benefit is subtler but real: velocity teaches you what an appropriately hard set actually feels like, which sharpens your judgement on lighter assistance work where you are not tracking speed at all. It is a feedback loop that makes every rep more honest — the same way a reliable strength-tracking baseline, as covered in AI strength testing and 1RM prediction, makes your programming more honest.
Train the body you have today, not the one from last month.
The AI Fit Blueprint folds velocity-based autoregulation into a complete system — adaptive overload, real-time load adjustment, and recovery-aware deloads — so every session lands in the productive zone instead of guessing from a stale percentage. Stop leaving gains on easy days and stop grinding yourself into the ground on hard ones.
Get the Blueprint →The Bottom Line
Fixed-percentage training is built on a fiction: that your strength is constant. It is not. It rises and falls with everything else in your life, which is why the same weight can feel light on Monday and crushing on Thursday. Bar velocity replaces that fiction with a measurement — a live, per-rep readout of how hard the set really is for you today.
AI makes velocity-based training accessible to anyone with a phone. It builds your personal load-velocity profile, prescribes loads that hit your target speed, caps velocity loss so fatigue stays inside budget, and flags the sessions where you should push or back off. The result is a program that adapts to the athlete rather than the calendar.
The rule is simple: let the bar tell you how much. Track speed, set a velocity floor, and let load float day to day. That is how you stop training blind and start training the body you actually have in front of you.