Your brain cycles even while you're awake
Nathaniel Kleitman, the University of Chicago physiologist who helped discover REM sleep back in 1953, made a second, quieter discovery about a decade later. The same roughly 90-minute cycle that drives your REM and non-REM sleep stages doesn't switch off when you wake up. It just keeps running.
He named it the Basic Rest-Activity Cycle, or BRAC. Your alertness, mental sharpness, and physical readiness rise and fall on a 90 to 100 minute cycle: about 90 minutes near the top, followed by roughly 20 minutes in a trough before the next climb starts.
The shape of one cycle

- Minutes 0-20: ramping up. Your engagement builds.
- Minutes 20-60: the peak. This is where your best work actually happens.
- Minutes 60-90: the tail. You can keep going, but quality is already sliding.
- Minutes 90-110: the trough. Capacity drops off sharply.
Push past minute 110 without disengaging and you're now running in deficit. The next "cycle" won't fully recover, because you never let the last one finish.
Sleep researcher Peretz Lavie, working at the Technion through the 1980s and 90s, confirmed the same roughly 90-minute swing across alertness, cognitive performance, EEG readings, and body temperature.

Why most workdays ignore this completely
The factory model never left. The eight-hour day was built around manual labor, where output stays roughly constant hour to hour. Knowledge work doesn't behave that way, but the schedule still assumes it does.
Pushing through feels good, briefly. There's a real dopamine hit in "I'm grinding," and caffeine is very good at masking the actual decline happening underneath it.
Your calendar doesn't care about your biology. A 2pm meeting isn't going to reschedule itself around your trough.
The cost of pushing past 90 minutes
Psychologist Anders Ericsson's research on deliberate practice turned up a striking pattern across elite performers in field after field: they almost never sustain more than about four hours of genuinely focused practice in a day, and that time comes in three or four chunks of roughly 90 minutes each, with real breaks in between.
A six-hour grind with no breaks produces less total useful output than three 90-minute cycles with genuine 20-minute breaks between them. That's not a motivational claim, it shows up directly in lab measurements.
What a real break requires

A real break means your executive-attention network actually goes offline:
- A walk outside, 10 to 20 minutes
- A short nap, 10 to 30 minutes
- Staring out a window, genuinely doing nothing
- A casual chat with someone about something other than work
What doesn't count:
- Checking email (still the same attention network, just a different target)
- Reading news that's adjacent to work
- Scrolling social media
A day, mapped out

A realistic structure:
- 8:30-10:00 Cycle 1: your hardest deep work of the day.
- 10:00-10:20 Trough: take the walk.
- 10:20-11:50 Cycle 2: second deep work block.
- 11:50-12:30 Trough, plus lunch.
- 12:30-14:00 Cycle 3.
- 14:00-14:20 Trough.
- 14:20-15:50 Cycle 4: reactive and collaborative work.
That's three to four real cycles, troughs actually protected, with the easier work pushed later in the day when your peak sharpness has already been spent.
What TaskCoach.AI does with this
Focus mode defaults to 25 or 50 minutes, and anyone who wants the full peak can set it to 90. Either way, the system enforces a genuine break after each block with a wind-down screen instead of letting you just barrel into the next task.
The mood and energy check-in builds your own personal ultradian profile over time. Most people are surprised to find their actual peaks don't match the cultural default of "mornings are for focus."
The Daily Clock view lays the day out in 90-minute slices so the structure is visible instead of theoretical.
The bottom line
Your brain runs in roughly 90-minute peaks separated by roughly 20-minute troughs, whether your calendar acknowledges that or not.
Three to four honest cycles a day is the real ceiling, which works out to 4 to 6 hours of genuinely high-quality output when it's structured well. The same total hours, run without protecting the troughs, get you maybe half the output and all of the burnout.
Work with the cycles instead of against them.