Your intervals end on a clock. Mine end when the work is actually done.
Every app prescribes '5 × 4 minutes.' Brevet ends the work rep when your modelled reserve hits the floor — and ends the set when you can't refill it. Here's why a stopwatch was always the wrong stopping rule.
2026-06-28 · 8 min read · Brevet · AI-assisted
Drafted with AI assistance and reviewed by a human. The science is sourced and real; the article isn't medical advice.
The short version: A normal interval workout ends each rep on a stopwatch — "four minutes, go again." The stopwatch knows nothing about you. Brevet ends the work rep when your modelled energy reserve drains to a floor, ends the recovery when it refills, and ends the whole set when you can no longer reach that floor — your real repeatable limit for the day. The clock was always a stand-in for the thing that actually matters. We went and read the thing.
The round number nobody questions
Open any training app and the hard day looks the same. Five times four minutes. Or six times three. Or forty-twenty. Clean blocks, clean clock, and you grind out the reps until the timer says stop.
I did those for years. Then one day, mid-set, lungs at the ceiling, the obvious question finally landed: why four minutes? Not three, not five, not "until the thing I'm trying to train has actually happened." Four because it fits a spreadsheet. Because it's a round number a coach writes down once and hands to ten thousand different riders.
Your physiology has never heard of four minutes. It doesn't round.
What the clock is standing in for
Here's what a hard interval is actually trying to do. Above a certain power — your critical power, the real line between "hard but steady" and "the tank is now draining" — you start spending a finite reserve. Sport scientists call it W′: the work you have above threshold before the bucket runs dry. Ride above that line and it drains. Drop below and it slowly refills. Monod and Scherrer started mapping this in the sixties; Skiba turned it into a balance you can model second by second.
That draining bucket is the stimulus. Not the minutes — the drain.
The four-minute block is a bet that, for an average rider, four minutes empties roughly the right amount of bucket. Sometimes the bet pays. Often it doesn't. A rider with a deep reserve sails through four minutes barely dented — a tempo block in a VO2 costume. A rider with a shallow one is cooked at two and spends the back half digging a hole that trains nothing they came for.
So we threw the clock out as the decider and kept it only as a guard rail.
End the rep when the work is done
In Brevet, a hard work interval doesn't end at a time. It ends when your modelled reserve drains to a floor. You hold the target, the battery falls, and the moment it touches the line — that's the rep. Sooner than you expect some days, later on others. Your drain rate decides, not a number a coach picked for a stranger.
Then the recovery runs the same trick in reverse. It doesn't end at a fixed rest. It ends when your reserve has refilled enough to go again — and how fast you refill is its own personal trait, slower if you barely back off, quicker if you genuinely soft-pedal. You get exactly the rest you need. Not the rest the template guessed.
I'll say the quiet part out loud. The drain is the stimulus, so the drain is the rule. Everything else is bookkeeping.
The set ends when you can't do it anymore
This is the part I'm proudest of, and it's almost embarrassingly simple once you see it.
A set should last exactly as long as you can keep hitting the quality. Not one rep longer. The reps after that aren't training the engine; they're central fatigue with no matching stimulus — the junk volume that makes you tired without making you faster.
So the set ends at the honest place: when you can no longer drain to the floor in the time you've got. Your reserve isn't coming back fast enough, your legs can't push it down to the line — that's your repeatable ceiling for today, and we stop there. Somewhere between a sensible minimum and a hard maximum number of reps, decided by you, on the day, in real time. A fresh, well-slept you finds a higher ceiling. A you who's three days into a heavy block finds a lower one, and we don't pretend otherwise.
When I rode it the first time, I didn't finish all the reps I'd have written down on paper. I ran out of repeatable floor a rep early. It felt exactly right — challenging, honest, done when I was done.
Anchored to your engine, not a slice of your test
One more thing the clock-and-percentage habit gets wrong: the target power itself. "Do your hard work at 120% of FTP" assumes everyone's reserve sits in the same ratio to their threshold. It doesn't. Two riders with the identical FTP can have very different reserves, different maximal aerobic power — the wattage where the aerobic engine genuinely tops out.
So the work isn't pegged to a flat slice of your test. It's anchored to your own maximal aerobic power, drawn from your critical power and your reserve, so the effort lands in the zone that drives the adaptation instead of a textbook number that's too soft for one rider and too sharp for the next. There's a deeper-draining variant for anaerobic days and a shallower one for VO2 days, because they chase different things.
I'm not going to print the floors and the anchor math here. That part took the longest to get right, and it's ours. The principle is the part worth sharing.
What I'm not claiming
W′ balance is a model, not a fuel gauge wired to your blood. It runs on an estimated critical power, and a model on a shaky estimate is just a confident guess. So the whole system stays switched off until you've earned it — a real, recent, all-out test. No test, no trustworthy threshold; no trustworthy threshold, no draining you to a floor. We fall back to the honest old clock and tell you why. I'd rather under-deliver the magic than titrate you against a number I don't believe.
And you're never a passenger. The target is always something you can simply hold; the floor is a recommendation, not a cage; and you can ride the planned block to the second if that's what you want. The physiology gets a vote. It doesn't get the wheel.
After the ride: see what your body actually did
When you're done, you don't get a "64% — underperformed" slap because your set ended at four reps instead of a planned five. That grade is backwards for a workout built to end when you do. Instead you see the trace: your reserve draining to the floor and recharging, rep by rep, the whole sawtooth you just rode. You see an estimate of your VO2max from the session. And you get the line I think is the honest summary of the whole idea — you found your repeatable ceiling today — with the number of reps it took to get there.
Watch that ceiling across a block of training and you have something a fixed "5 × 4" can never give you: a clean, personal readout of whether your engine is getting deeper, or whether you've been digging too hard and need to back off. The workout stops being a box to tick and becomes a measurement.
Why nobody else does this
Zwift plays a video. TrainerRoad plays a very good fixed file. Garmin hands you a structured workout off the same clock. They'll let you skip or extend a rep by hand — but the machine is still counting seconds, because driving a trainer off a live, individual reserve model is genuinely hard, and getting it honest enough to trust is harder.
That's the whole game. Not "more science" as a slogan — a stopping rule that finally matches what the science has said for forty years: above threshold, the drain is the dose, and the dose is personal.
The clock had a good run. It's just not the coach.
A note on what this is. Brevet is a training and coaching tool, not a medical device. W′ balance, critical power, and the VO2max estimate here are training concepts — a model of your performance built from your own test and ride data, described against your own history, never a health assessment. The system recommends; you decide every rep, and the planned target is always yours to hold.