Deep work: the ninety minutes have to exist first
It is 10:15 and you have been at the desk since nine. In that time: two threads on Slack, one call that was going to be quick, an alert for a meeting you are not in, and the document you opened at 9:02, still on the paragraph you opened it at.
You did not slack off once. You were available the entire time, which is the same thing in better clothes.
The morning was not lost to weakness. There was never an uninterrupted ninety minutes in it to lose.
Ninety uninterrupted minutes is a scheduling problem before it is a discipline problem
This is the whole article, so it goes first.
People who cannot do deep work usually assume the missing ingredient is willpower — that other people simply want it more at minute forty. Almost always the block does not exist. It was never written anywhere, nobody else knew about it, and at 9:20 something perfectly reasonable walked into it because the space was unclaimed.
A block you have to defend in real time will lose, every time, because defending it costs more attention than the work does. A block that already exists on the calendar with a name on it costs nothing to defend, because the argument happened last week.
Put the block on the calendar first. Everything below is detail.
What if forty minutes is all I have?
Then run forty. A real, protected, named forty-minute block is worth more than a theoretical ninety you keep meaning to arrange. The reason the number ninety keeps appearing is not that shorter blocks fail — it is that long-warm-up work spends its first fifteen minutes just getting the problem back into your head.
What Cal Newport actually argued
Deep Work came out in 2016. Newport is a computer science professor at Georgetown, and his argument is closer to economics than to self-improvement: the ability to concentrate hard on a difficult thing for a long stretch is becoming rarer at the same time as it is becoming more valuable, and that gap is a career opportunity for anyone who keeps the skill.
He splits work in two. Deep work is cognitively demanding and done without distraction. Shallow work is logistical, easy to do while distracted, and expands to fill whatever you give it — email, scheduling, status updates, most meetings.
Worth being honest about the evidence: the book is an argument built from case studies, history and reasoning, not from controlled trials. Nobody has run the experiment where half a profession works in ninety-minute blocks for two years. Take it as a well-made case rather than a measured result — and notice that it is useful mainly because the shallow half of the description is instantly recognisable.
The finding that most supports it comes from elsewhere. Sophie Leroy’s attention residue work (Organizational Behavior and Human Decision Processes, 2009) found that leaving a task unfinished to switch to another leaves part of your attention behind, measurably hurting the second task. Short lab tasks, so the effect size is unsettled, but the direction explains why a morning of nine small things feels like a morning of nothing.
How it differs from Pomodoro
They solve opposite problems, which is why arguing about which is better goes nowhere.
| Pomodoro | Deep work | |
|---|---|---|
| The problem it fixes | Starting | Being interrupted |
| Session length | 25 min, fixed | 50–90 min, chosen per block |
| What ends it | A bell | A milestone you named beforehand |
| What it costs to set up | Nothing | A conversation with your calendar |
| Best for | Admin, revision, avoidance, shapeless days | Code, drafting, design, analysis |
If your problem is that you cannot get going at all, deep work is the wrong first move. Go and make starting cheap for a fortnight, then come back. Nobody builds a ninety-minute habit on top of a starting problem.
Building the room
The block is time. The room is everything else, and it takes about ten minutes to build once.
One screen, one document. Not one window with a second one peeking. Full screen, the thing you named, nothing else reachable in one gesture.
The phone unreachable, not face-down. A drawer, another room, or blocked at the system level for the length of the block. Distance buys you seconds and seconds are what this is made of.
One sound, every time. Rain, brown noise, a café loop, or silence — the specific choice matters much less than using the same one, because after two weeks the sound itself becomes the signal that the block has started. Which sound and why is a longer argument, and the short version is: nothing with words in it.
Something to write stray thoughts on. Paper. Every idea that arrives mid-block goes there in four words and gets dealt with afterwards. The thought’s actual demand was not to be acted on, it was to not be forgotten.
Do I have to quit social media?
No. Newport argues for a stricter relationship with these tools than most people want, and you can decline that argument entirely and still run a block. All the block requires is that the apps are unreachable for eighty minutes. Whether you delete them at 6pm is a separate conversation and not one this article is having.
A start ritual, under two minutes
Same order, every time, until it stops being a decision.
- Close everything except the one document. Twenty seconds.
- Write the milestone on paper. Not “work on the report” — “the method section has a rough first draft”. Forty seconds.
- Phone away, sound on, timer started. Fifteen seconds.
- Write one bad sentence, or run one broken test, inside the first minute.
Step four is the one that carries the ritual. Beginning with something deliberately poor removes the only decision that reliably kills the first ten minutes: what the good version would look like.
The block: a milestone, not a bell
A bell says the time is up. A milestone says the thing you named is done. They fail differently, and the difference matters for long work.
With a bell, arriving at minute 88 with two paragraphs to go turns into a negotiation you have twelve more times before you finish. With a milestone, you decided in advance what “done for now” means, so at minute 88 there is nothing to negotiate — you either have it or you carry on until you do.
Two rules keep this honest. The milestone has to be small enough that you hit it most days, or it is a wish rather than a target. And you need a real stop somewhere: a meeting, lunch, the school run, something with its own gravity. A block with no outer edge at all becomes the failure mode that Flowtime people know well — three good hours followed by a dead afternoon and a bad next morning.
How long should a block be?
Fifty minutes if you are new to this or the day is fragmented. Eighty to ninety once the ritual is automatic and the calendar cooperates. Beyond about two hours most people are paying for the extra time out of the following day’s account.
You will see ninety minutes explained by biological cycles. That argument exists and has more caveats than the internet usually mentions — the evidence is much stronger for what happens during sleep than for what happens at your desk.
A starter week
The first mistake is scheduling five. Here is a week that survives contact with an actual job.
| Day | Block | Length | Why |
|---|---|---|---|
| Mon | 9:00–10:20 | 80 min | The least contested slot most people own. Defend this one hardest |
| Tue | — | — | Meetings. Do not fight Tuesday in week one |
| Wed | 9:00–10:20 | 80 min | Same slot as Monday. The sameness is what makes it survive |
| Thu | 16:00–16:50 | 50 min | A short one at a different energy. Keep the milestone small |
| Fri | 9:00–9:50 | 50 min | For finishing something, not starting it |
Four blocks. Four hours and twenty minutes for the week. That is a real deep work practice, and it is roughly four hours more than the morning described at the top of this page.
Two more things make it hold. Put it in the calendar as a busy block with a name, so it looks like a commitment to everyone including you. And tell one person — the colleague most likely to interrupt you — that you are unreachable then and reachable at 10:30.
Most requests survive eighty minutes. The ones that do not are allowed to break the block; nothing is being scored.
When this is the wrong tool
Being reachable is the job. Support, on-call, clinical work, anything with a queue. The honest version for you is one protected block a week, not a daily practice, and pretending otherwise just adds guilt to a job that has enough.
You share the room and cannot change that. Then build the time instead of the space: 7:30am at the kitchen table is a better block than 2pm in a room with three other people in it.
You are in a bad stretch. Illness, a newborn, a genuine overload. Long blocks are the first thing to go and should be. Twenty-five minute sessions still work when eighty do not, which is most of why they exist.
Fomodo has a Deep Work mode: you name a target, and when you reach it the app tells you and keeps counting rather than stopping you, because the end of a good block is rarely the end of a round number. It will shield your distracting apps for the length of the block and keep the log afterwards.
The thing it cannot do is the thing this article is about. It cannot put the block on your calendar, decline the 9:30, or tell your colleague you are unreachable until 10:20. It guards the edges of a block that already exists. Making it exist is the part nobody can automate for you.
The block first. The ritual second. The app is somewhere around fifth, and honestly the paper matters more.
