What forty minutes a night costs you by March

It is later than you meant it to be, the room is dark, the phone is the only lit thing in it, and you have already told yourself once tonight that you were putting it down.

This piece is arithmetic. Not advice about winding down, not a lecture about blue light, just a number you can derive from data your phone already keeps and multiply out to a date on the calendar. The number is yours, not ours, and you will have it in ten minutes. Then the harder part: what the evidence says a deficit that size does, and why you were never going to detect it by how you felt in the morning.

First, get the honest version of your nightly overshoot

The temptation is to guess. Do not guess. Your phone logged it.

1. Fix your wake time first. Whatever your alarm says. This is the number that makes the subtraction real, because if your morning is fixed, every minute you spend awake at night comes out of sleep and nowhere else.
2. Write down your intended lights-out. Not an aspiration. The time you would need to be asleep to get the sleep you think you need before that alarm.
3. Open Settings > Screen Time > See All App & Website Activity and stay on the Day tab. The chart runs across the hours of the day. What you want is the far right of it, the last hour with anything in it at all.
4. Subtract. Last use minus intended lights-out. That is one night's overshoot.
5. Do it for seven days, then take the middle value, not the average. One catastrophic Saturday drags an average somewhere useless. The median is the night you actually have.

Two fences around that number, both of which make it smaller and both of which are honest. Some of those minutes you would have spent lying there awake anyway. And if your wake time floats, if you claw back an hour on Sunday, the debt really is smaller than the subtraction suggests. Take the fences. The number survives them.

Now multiply it by the number of nights between here and spring

There are five nights left in August. Then September has 30, October 31, November 30, December 31, January 31, and February 28, which is 181. Five plus 181 is 186 nights between tonight and the first of March.

The arithmetic is one line. Your nightly minutes, times 186, divided by 60 for hours, divided again by 8 for whole nights of sleep. Here it is worked for five plausible numbers so you can find yours.

Forty minutes a night is a hundred and twenty four hours by March, which is fifteen and a half whole nights of sleep that will not happen.

Forty minutes is the modest answer. It is the amount of time that feels like nothing at the moment you spend it, the amount you would not bother to defend, the amount that fits inside the phrase "just a bit longer". It is also two full working weeks of nights.

The total is a good headline and a bad model of what is happening

Here is where most versions of this argument overreach, so this one will not. You are not going to arrive in March owing fifteen nights in the way you might owe fifteen pounds, and nobody is going to hand them back to you. Sleep debt is not a ledger with a balance. The total is useful only because it makes the size of the habit visible. What actually matters is the nightly deficit, repeated, and what a repeated nightly deficit does.

For that there is one study worth knowing about. In 2003, Van Dongen and colleagues put 48 healthy adults, aged 21 to 38, on 4, 6 or 8 hours in bed for 14 consecutive nights, with a separate group kept awake for three nights straight as a comparison. Restriction to 4 or 6 hours produced significant, cumulative, dose-dependent deficits on every cognitive task they measured. Cumulative means the damage kept accruing across the fortnight rather than levelling off after a few bad nights. Dose-dependent means the size of the cut predicted the size of the deficit. Two weeks at six hours a night left people with vigilance deficits comparable to about two nights of no sleep at all.

Be careful with the transfer, because it is easy to abuse. Six hours instead of eight is a two hour cut, and forty minutes is a third of that. We have no figure for what a third of that dose does over 186 nights, and this piece is not going to invent one. What does transfer is the shape of the finding: the deficit accumulated rather than settling, and it scaled with the cut. That is enough to make your own number worth knowing.

The reason you have never noticed is in the same paper

The result from Van Dongen that matters most, and the one that makes this an article about arithmetic rather than an article about willpower, is this. The participants' own sleepiness ratings did not track their objective decline. People rated themselves only slightly sleepy while performing markedly worse.

Read that again in bed. Your internal readout is not a reliable instrument for this. If the deficit announced itself, if forty minutes felt like forty minutes of damage the next afternoon, you would have fixed this years ago without needing a table. It does not announce itself. It shows up as a slightly worse meeting, a slightly shorter fuse, a reread paragraph, and every one of those has a more available explanation than last night's phone.

The mechanism is the hours, not the light

You have probably already tried the optical fix. Night Shift is likely on. Maybe you bought the glasses.

In 2024 a National Sleep Foundation expert consensus panel reviewed screen use and sleep across the lifespan and, for adults, did not reach consensus that screen use impairs sleep health, and did not reach consensus that blue light from screens before bed impairs it. For children aged 5 to 12 they did reach consensus, both on screen use in general and specifically on the content of what is watched. And they did reach consensus that behavioural strategies can reduce the negative effects.

The glasses have been tested directly. A meta-analysis of double-blind crossover randomised trials found blue-light blocking glasses shortened sleep onset latency by 4.86 minutes, which was not statistically significant, with no effect on sleep efficiency or waking in the night. That was three trials and 49 people in total, which is small, and not significant is not the same as proven useless. But note the size of it against your own number. Even taking 4.86 minutes at face value, it recovers about a ninth of a forty minute overshoot.

What is left, once the optics are set aside, is unglamorous and much harder to argue with. The hours are simply spent awake. That needs no study; it is the subtraction you did at the top. The second mechanism is the content: fear, anger, excitement and anticipation delay sleep, which is the pathway the NSF panel endorsed in children and the plausible one in adults. It is not the light coming out of the phone. It is what the phone is for.

Which is why the thing you are doing right now has a name people reach for. "Revenge bedtime procrastination" translates the Chinese 報復性熬夜, retaliatory staying up late, which appeared on Weibo around 2016 in response to 996 working culture and reached English through a tweet by the journalist Daphne K. Lee in June 2020: people with little control over their days refusing to sleep, to reclaim some freedom at night. It is a popular coinage, not a diagnosis. It describes the feeling well, explains nothing, and will not give you back the 124 hours.

What a rule has to do, given all that

Everything above points one direction. The cost is real and it accrues, you cannot detect it from the inside, and no filter touches it because the mechanism is time rather than wavelength. So the decision has to be made once, in daylight, by someone who has seen the table, and then it has to hold at midnight against someone who has not.

That is the whole design brief for Scrappy. He is a small scruffy dog who takes one job seriously: quiet hours you set while sober, enforced when you are not. His shield escalates when you push, from polite to surprised to firm to a flat "No." to genuinely wounded. If you want minutes back you walk for them, roughly 2,000 steps for one treat, and a treat buys five minutes. Treats are not sold and never will be. Deleting the app is the obvious way out, so there is a removal lock, and switching it off takes a full day, deliberately longer than an urge lasts. There is one emergency pass a month, no scolding, because some nights really are emergencies. No account, no sign-up, no server of ours; Screen Time is drawn by a sandboxed Apple extension that cannot phone home. The free tier is one complete, uncrippled rule plus every safety rail, which for this problem is the only rule you need.