Ninety Minutes With the Door Shut: Cal Newport on Deep Work, Email, and a Life Built for the Worst Case

Cal Newport is a computer scientist at Georgetown University whose academic specialty is the theory of distributed algorithms, including impossibility results: proofs that some problems cannot be solved at all, or cannot be solved faster than a given bound. A much larger audience knows him as the author of *Deep Work* and *Digital Minimalism*, and, at the time of this conversation, a new book called *A World Without Email*. He does not use social media. He hosts a podcast, *Deep Questions*, aimed at people who want a more deliberate working life. Lex Fridman, who has used Newport’s writing as a guide, opens with the obvious question: what is deep work?

Newport’s definition is plain. It is his term for focusing without distraction on a cognitively demanding task — something everyone has done, but had not usually named as a category of effort separate from the rest of the workday. He had been calling it “hard focus” on his blog. The idea hardened while he was in the theory group at MIT’s CSAIL, surrounded by what he calls a murderers’ row of thinkers: Turing Award winners, MacArthur fellows, people for whom the ability to concentrate was a tier-one skill, discussed the way athletes discuss a 440 time. A typical day, he says, might begin with Eric Demaine at a whiteboard, a graph-theory grid projected on it, visitors clustered around. You went to lunch, you went to the gym, you came back. They were still staring at the same board.

Fridman admits that hanging out with mathematicians and physicists can make experimental machine learning feel like a trick. Newport’s reply is that original algorithmic work, and even hard coding, is a classic case of deep work: you have a problem, a pile of tools, and you have to combine them on the fly. After *So Good They Can’t Ignore You* came out in 2012, readers asked how to get that good. He looked back at graduate school and wondered how general the focus habit was. The hypothesis that followed is the spine of the book: across knowledge work, concentration matters, and we have accidentally built circumstances in which we barely do it.

The keyword in Newport’s definition, he insists, is *without distraction*. A big theme of his work is that context shifting kills the human capacity to think. Even a brief look away, then a return to the main task, produces what he calls a cognitive pile-up. Neural networks that were supporting the original work have to be suppressed; others have to be amplified. Rip yourself out halfway through that cascade, then try to switch back, and your ability to think clearly drops. Do it long enough and by midday you are not tired in the ordinary sense. You simply cannot think anymore.

Fridman asks the programmer’s question. You are inside a function, then you wonder whether a library already does this, and you go to the internet. Is that a full context shift, and should you always refuse it? Newport is milder than the folklore around his book. Looking up a relevant library is probably fine, because the semantic networks are similar: you are still thinking about this kind of solution, this kind of function. The hit comes when you switch to something different *and* leave an unresolved obligation. The worst case is the email inbox: twenty messages you cannot answer now, in a totally different context, with someone waiting on you. The second worst is emotionally arousing content. Glance at Twitter, he says, and you can wreck the cognitive plateau you were standing on.

He wants blocks of at least an hour, because it can take up to twenty minutes to clear the residue of whatever you were thinking about before. Code for thirty minutes after a switch and you may get only ten or fifteen minutes of actual peak work. An hour buys you forty or forty-five. He is partial to ninety minutes as a chunk that gets a lot done just before exhaustion, at which point you can pull back. Fridman adds a personal report the books do not need: the days he gets several hours of that kind of work, he is not only productive. He is less of a jerk to the people around him. The days he does not, irritability and a sense of having accomplished nothing pile up until the rest of life is hard to enjoy.

Asked how to put deep work into an actual week, Newport does not start with willpower. He starts with time blocking. Do not let the inbox or the to-do list drive the day. Look at the time you actually have — this meeting, that appointment, the gaps — and assign it. A half hour here, a ninety-minute block there, five errands packed into a gap between meetings. Once you are in that discipline, it becomes easier to see where the deep work should sit and to park shallower obligations somewhere they do less damage. You will get more of the hard thing done than if you arrive and ask, “What’s next?”

He plans on three nested scales. At the quarter or semester, a big-picture vision: a paper deadline at the end of the season, a number of book chapters. Weekly, he looks at that vision against the actual calendar and decides which days can carry the load — maybe Monday is the only empty day, so that is the writing day. Daily, he takes the weekly plan and blocks the hours. Fridman, who religiously avoids meetings and treats a ten-minute Zoom as an existential crisis, often schedules a focus hour and hopes to get lost and overrun it. Newport shares the instinct. If a deep work session is going well, he says, “rock and roll and let it go on.”

The planner he sells has many columns for a reason. When the initial schedule breaks, you move one column over and fix the plan for the time that remains. There is no bonus for having stuck to the first draft. The prize is an intentional plan: at any point in the day you have thought about what time remains and what to do with it. He will take the yelling that comes from protecting a proof when the pens start to click. Fridman notes the other side of the same coin. Stopping while things are going well can make you eager to return; stopping in a dead end adds a force of procrastination the next time you sit down.

Fridman turns the conversation into a therapy session: it seems he only finishes things that have deadlines. Why do they work so well, and can anything get done without them? Newport’s short-term answer is that a deadline is a clear motivational signal. Time blocking does a similar job at a smaller scale. If you merely say you will write for a while, your mind knows you will take breaks eventually, so every three minutes it proposes one. On a blocked schedule you have two hours for writing and a break already placed somewhere else. You do not have to fight yourself. At a larger scale, a due date tells you what you are supposed to be working on. The healthier move, in *A World Without Email* and in the rest of his writing, is to build a process and then give yourself over to it.

He does not like the other kind of pressure — the friend noting that you promised this two months ago — and he tries to finish early. He wants flex. The night-before crunch puts a thought in his head: what if I get sick, what if I sleep badly and cannot think? Aristotle, he notes, ended up locating the human good in the capacity to contemplate; Newport tries to train a cognitive habit the way people train a fitness habit, until skipping it feels off. Fridman describes the opposite pattern: successful processes for deep work, then extra chaos piled on top, too many interesting things, not enough sleep. Newport’s old website motto was “do less, do better, know why” — a few bets on the roulette table, trained hard, tied to a value, and then you see where they lead. Passion, as he argued in 2012, is cultivated as skill and autonomy grow. It is not a torch you are supposed to find fully lit before you start.

He learned to shut the day down early because he married early, while still a graduate student, and treated his wife’s office hours as a hard stop. As a postdoc he put artificial constraints in the middle of easy days — two hours of exercise, productive meditation — so he would not get flabby before the professor job arrived. Kids, he says, are a huge productivity hit and also the ultimate timetable enforcer. He worries about burnout the way he once worried about “deep procrastination” in elite students: a hard load plus a motivation that is too extrinsic, and the system fries. His counter is seasonality at every scale — hard weeks and easy weeks, hard semesters and easier ones, a lighter summer, a five o’clock cutoff — rest and recovery as the point, not a reward you earn after collapse.

In *Deep Work* there is a chapter called “Embrace Boredom,” and Newport now calls its original argument functionalist. You have to keep some boredom in the regular schedule or the mind forms a Pavlovian link: the first hint of understimulation, and it demands a hit. Once that link is in place, it will never tolerate three hours of writing. Lately he is after something deeper. Boredom is extremely uncomfortable. Hunger, thirst, and pain are extremely uncomfortable because they serve the species. So what is boredom for?

His working theory is that boredom drives productive action. What got humans to work with fire, to shape hand axes, to sharpen a stick until it could be thrown at a mammoth? Boredom, pushing toward doing. Then the analogy: we have a hunger instinct, and industrial food built hyper-palatable junk that short-circuits it and makes us unhealthy. We have an action instinct, and the modern environment overloads us with tasks and distractions until the system shorts — more on the plate than anyone can finish, then frustration that we cannot finish it. He calls the result a kind of cognitive action obesity. The open question, which he is still thinking through, is the paleo version for action: if you could build from scratch how much to do and what to work on, what would put you back in touch with that drive?

Fridman takes the analogy into the body. He has been eating one meal a day, mostly meat, for more than a month. Fasting, even a single day, put him in a listening mode: hunger is a signal, and different foods answer it differently. Two thousand calories of steak with some green beans or cauliflower leaves him fulfilled, focused, and happy for hours. Two thousand calories of cherries, a food he does not know how to moderate, leaves him bloated, mood-swinging, hungrier the next day. Fasting taught him to hear the signal instead of pacifying it. That, both of them agree, is the same move as embracing boredom: take the hyper-palatable filler away and notice where the discomfort is actually pushing you.

In *Digital Minimalism* Newport asked readers to spend thirty days away from digital personal entertainment: social media, online video, anything that captures attention and dispels boredom. People heard “detox.” That was not the point. He wanted space in which you could listen to boredom and ask what it was pushing you toward, with the new screens taken off the table. About seventeen hundred people went through the experiment. It was hard at first. Then they got used to listening. Boredom pushed them toward other people, toward high-quality leisure, toward something a little complicated. The library came up a lot: physical books, low pressure, a window, weather. I used to do that twenty years ago, they told him.

Digital minimalism, as he defines it, works backwards from what you actually want to spend time doing, then asks whether a tool can amplify that. After the month you bring technology back only in that role. Abstention-only plans — “I am on Instagram too much, so I will use it less” — fail in the moment, because in the moment it is not that bad. A positive picture of a life is much stickier. The people who made lasting changes were the ones who experimented and reflected. The people who just white-knuckled thirty days because they were sick of their phones did not last. Roughly half of those who wrote to him dropped social media altogether. The other half kept some of it, but almost none of them kept it on their phones. Once you know the function you are optimizing, you can write rules, and the cost-benefit ratio swings from the company’s side toward yours.

Fridman has been running a smaller version of the same experiment with Twitter. Posting once or twice a day and checking around those posts feels good: negative comments slide by, positive ones make the world seem full of decent people. Somewhere around five or six checks a day the valence flips. Negatives stick, positives feel shallow, and what remains is not even anger so much as an existential exhaustion — a sense of not having made anything beautiful. Newport has seen many configurations. For people with a public, a single daily post can support a real value, informal contact with an audience, if you fence it. His writer friend Ryan Holiday posts one Stoic quote a day and leaves. He never interacts. Readers like the dose of inspiration. That is the pattern: decide what matters, use the tool to amplify it, and stop living inside the company’s loop.

Newport’s unified theory of social media is that he is very bearish on the big platforms having a long future. Their original pitch, especially Facebook and Instagram, was that the people you already know are here. That is a network-effect moat: value grows sharply with users, and a latecomer cannot catch up. When they shifted toward a news-feed model — we will show you interesting stuff, we will chase likes — they stopped being a social tool in that sense. The impulse to connect digitally moved to group texts and WhatsApp. A feed that is merely diverting has to compete with everything else that is diverting, including podcasts. There is no more centrifugal force. “Is a Twitter feed really that much more interesting than all these streaming services?” he asks. “Is it really that much more interesting than Clubhouse? Than podcasts? I feel like they don’t realize how unstable their ground actually is.”

His prediction is fragmentation into what he calls long-tail social media. If you do not need everyone you know on one platform, why not three or four bespoke ones — a thousand people who care about AI or comedy, an interface perfected for that group, maybe audio, two dollars a head so no one has to harvest attention? Comedians, he notes, are already on podcasts; Twitter was never the best format for them. The internet’s old democratic power is still a feature. What Twitter adds is homogenization: everything looks the same and moves with the same low friction, which destroys distributed curation. In the blog era, people owned servers and reputation accumulated through time and links. During the 2004 election you had Daily Kos, the Drudge Report, Ezra Klein writing from a dorm room. A conspiracy blog had a hard time getting on anyone’s radar. On a fully homogenized, low-friction network, almost anything can shoot through.

Early internet people, the ones who were on servers before the web, were flabbergasted by the rise of the platforms. They had built open protocols so any machine could talk to any machine. Then companies said they would own the servers, put them in buildings they owned, and treat the internet as the first mile into a private internet. Jack Dorsey, Fridman notes, has been funding more distributed versions of Twitter, which could in principle destroy the product. Newport’s retort is that the tech is not the point. Bike enthusiasts will jump on Zoom, keep a Slack, put up a blog. Meanwhile engineers want a beautiful federated protocol nobody understands. Clubhouse, in its early form, delighted him as single-use technology: mute, leave quietly, almost nothing else. Bits are cheap enough now that not every community needs a unicorn story about a hundred million in revenue. Facebook was a fantastic capital-concentration machine, which is a hard thing for a club to fight. Then again, WallStreetBets showed that a lot of small clubs can team up.

The real title of *A World Without Email*, Newport jokes, should have been *A World Without the Hyperactive Hive Mind Workflow*, but the publisher wanted something pithy. He started the reporting right after *Deep Work*, with interviews in 2016, and spent five years pulling the threads. The question was simple: if deep work is so valuable, why does almost nobody have time for it? Why the constant check of email and Slack? Email spread for boring, good reasons. Between 1990 and 1995 it made its move, replacing the fax machine, voicemail, and the paper memo. It was a killer app because it was useful. In its wake came a new way of collaborating: we will just work things out on the fly with unscheduled back-and-forth messages. That workflow followed email the way, in his plague metaphor, the virus followed the rats. Today the average knowledge worker checks one of these channels every six minutes, because every relationship in the organization — colleagues, vendors, clients, HR — now runs on asynchronous, unscheduled ping-pong.

Nobody paused to design a system that would actually work. The change was emergent, a case of technological determinism: the mere presence of a tool can change how people act. In the mid-to-late 1980s, IBM’s Armonk headquarters prepared to put in internal email. Mainframe time was expensive, so they provisioned from a study of existing communication — memos, calls, notes. Newport spoke with the engineer who ran it. Three days after launch, the machine had melted down. People were communicating six times more than the estimate. That is not enough time for an all-hands meeting announcing a new culture of over-communication. Low-friction digital messaging, left on the table, rewrote collaboration by itself.

Large offices had a real problem that email solved: fast asynchronous communication. Synchronous calls had too much overhead; memos were too slow. When the CIA built its headquarters, the need was so sharp that they installed a pneumatic-tube “email” system, Plexiglas capsules, brass dials, electromagnetic routers, a message arriving a floor away in about a minute. That kind of money is a measure of how badly the institution wanted speed without a meeting. So when actual email arrived it was a productivity silver bullet. Researchers building custom computer-supported collaboration tools in the late 1980s watched the whole research program get wiped off the map. Then the unintended consequence: a miracle application that also blew up the immune system around attention.

Newport’s contention is that the hive mind helped at first and then became so costly that it has been pulling industrial productivity down. The metric has stagnated for a long time. The only reason it has not fallen, he thinks, is off-the-books extra shifts: three hours in the morning, three at night. We built a worldwide high-speed wireless internet and put supercomputers in our pockets. Productivity should have shot off the charts. It did not, because a brain cannot context-switch every six minutes. Context switching is the poison. We have been treating the symptom as a personal failure of will or etiquette. He will exonerate the tool. He would rather use POP3 than a fax protocol. The issue is the workflow. Collaborate with twenty or thirty people by unscheduled messages and you are standing at thirty metaphorical ping-pong tables. When the ball comes back you have to hit it soon or work grinds to a halt. You cannot batch, you cannot take a real break, you cannot build a better habit, because delay slows everyone else down.

The fix is the one the industrial sector already made: process engineering. Identify the repeated processes that make up the day — answering client questions, posting a podcast episode — most of which are currently implemented as “we’ll figure it out on Slack.” For each one, ask whether there is a way that minimizes unscheduled back-and-forth. You still send a file by email. You stop coordinating by email. Slack, in this telling, is just a faster hive mind and does not solve the thing. An anonymous well-known CEO told him this would be the moonshot of the twenty-first century, hundreds of billions in latent productivity. Tools can help — Trello, Basecamp, Asana, scheduling software — but the question is not “what replaces email.” It is “what process reduces messages,” and only then “what tool helps.” Some of those processes take more time in the moment. That is fine. The killer is the back-and-forth, because it forces the check.

Most of Newport’s interview booking for the book ran through a shared document with his publicist. She dropped options in; he checked twice a week and marked what worked. It takes longer than a quick email. It produces almost no back-and-forth. With his UK publisher, where the volume did not seem to justify a process, he felt the difference immediately. Fridman reaches for Shannon: spend time up front on a clever code, and you use much less bandwidth later. A half day of process can save hundreds of emails over six months. The trap on the other side is over-optimization. Software teams already have Scrum and kanban, which beat the hive mind, and some programmers still get obsessive — a level-three scrum master making the ritual seven times more complex than it needs to be. Newport’s piece “The Rise and Fall of Getting Things Done” followed Merlin Mann and the early-2000s “productivity porn” scene, people convinced that the right software configuration could offload the difficulty of work. The movement fell apart because work is hard, and deciding what to work on is hard, and no system does that for you. Get the process good enough to kill the context switches, then get after it.

When the question becomes what email is *for*, Newport is clear. It is good at delivering information. Think of it as a very good fax machine: send a file, broadcast a policy. It is bad at collaboration. If you are having a conversation, trying to reach a decision, trying to learn or clarify something that is more than a one-answer question, you should not be doing it in email. Novel messages still belong there — someone you have not heard from in a while, a note over the transom. That is email at its best, the *You’ve Got Mail* version. Where it starts to kill people is when all of the repeatable collaboration of a business has moved into the same channel.

For people stuck on a hive-mind team, he recommends asymmetric optimization first. Change what you can control, and be socially careful. Do not announce a new process everyone must follow. Say you will drop a draft in the Dropbox folder by noon Monday, leave it until Tuesday morning, look at comments then, and keep office hours Tuesday afternoon if anything is blocked. The other person hears that Cal has a plan and knows when to act. You have actually pulled them into a process that does not require a dozen replies. Do that a few times, then you can start the conversation — and, he adds, maybe slide the book across the table.

Fridman mentions the comedian Joey Diaz, who just picks up the phone. It makes Fridman uncomfortable. It also gets things done. Newport’s computer-science answer is that moving from synchronous to asynchronous networks adds a huge amount of overhead; the protocols get more complicated and more fragile. Real time is usually better. The human voice also carries more social information. It is generational: their generation will talk on the phone in a way a younger one will not. With close friends he does not cold-call. He schedules, and at the end of each conversation they take a couple of minutes to set the next one, the way you book a haircut so that you actually get one. Digital minimalism taught him a related paradox: the more you sacrifice nontrivial time and attention on behalf of someone, the stronger the connection feels. Social media took the friction out, and the brain stopped counting it as a real bond.

The conversation finally turns to the research that trained Newport’s taste for focus. What he has always liked in theory is impossibility: within distributed algorithms, his specialty is results that say nothing exists which solves this, or nothing exists which solves it faster than this. The lineage runs to Turing’s 1936 paper on computable numbers. Programs are finite, hence countable; problems, represented as infinite binary strings, are uncountable by Cantor’s diagonal argument. Vastly more problems than algorithms, so almost anything you can name is almost certainly not solvable by a computer. Then Turing gave a concrete case, the halting problem, and theoretical computer science has been in that business ever since. Impossibility proofs, unlike the analysis of a particular algorithm, are statements about the problem itself. They proceed by reductio: assume a solution exists, crank the logic until the universe blows up, retract the assumption.

Fridman reads from a paper of Newport’s on smoothed analysis of dynamic networks. Worst-case thinking often yields extremely strong lower bounds. The question is whether the bound is robust — capturing a real difficulty introduced by dynamism — or fragile, depending on an exact sequence of adversarial changes. Fragile bounds leave room for algorithms that still work in practice. Smoothed analysis is not Newport’s invention. Spielman and Teng introduced it for sequential algorithms, in particular the simplex method, which was fast in practice and could be terrible on the right pathological input. They let an adversary choose the input, then added a little random noise to each number, and simplex became fast. The “it can run forever” bound was fragile. Newport and collaborators carried the idea into distributed algorithms, where many machines talk across a network whose links can change every round. If a bad guy is driving those changes, strong lower bounds are easy to prove.

A paper they had just submitted looked at a simple rumor-spreading problem: some messages are in the network, and everyone should get them. If the topology is static, you can pipeline well. If it can change every round, existing lower bounds say it takes a long time no matter what algorithm you write. Smoothed analysis asks whether that is just extremely bad luck. They let the adversary still pick a worst-case network, then randomly add or remove an edge or two each round. The simplest possible solution blew past the lower bound. It is almost impossible, in that noisy world, to keep things slow. In the culture of theory this can be a problem of another kind: a result that falls out too easily may be rejected for insufficient “math machismo,” not enough mathematical self-flagellation. Newport is more interested in how many famous distributed lower bounds would melt under a little noise.

On deep learning he sides with Max Tegmark: inscrutability is not fundamental to the models. Practitioners do not lie awake about existential risk, he says, because the daily reality is multiplying matrices, tweaking constants, and hoping the classifier’s accuracy clears a threshold before the deadline. Once the right mathematical language shows up, he expects elegant equations, then plumbing: a pattern recognizer here, a noise module there, AI as an engineering discipline. Fridman objects that huge, densely connected, dynamic networks may resist clean representations the way Wolfram’s cellular automata do, turning theorists into experimental biologists. Newport does not pretend to know. He has dabbled in biological algorithms — ant colonies explained by a distributed procedure rather than a physics equation — and he keeps finding that work on computers was obfuscating older math: Shannon, entropy, a 1990s combinatorics result sitting under a pile of contention-resolution bounds. His guess about P versus NP is in that spirit. Someone very smart will find another representation of what the machines are doing, and an existing theorem will fall out. He is in the discovery camp about mathematics, with the caveat that brains also take huge pleasure in seeing simplicity, so we may be built to find it even when it is not quite there.

Near the end Fridman asks the question the productivity literature usually dodges. How do love, friendship, and family fit a high-output life? Is there a tension, or can relationships charge the whole process? Newport does not treat it as a tradeoff to be optimized. Relationships are the whole deal. If he were giving a commencement speech, the question he would put to young people is this: bad things you do not control are going to happen — illness, collapse, a cratered economy, a death. What is the plan? Relationships are the buffer, because we are wired for them. Researching *Digital Minimalism* sent him down a rabbit hole on the brain and sociality. Almost all of the machinery, he came to think, is there to service connection. In a famine, the people you are tied to are how you do not starve. Social rejection lights up pain centers; missing someone hurts for a reason. A widely accepted view of the default-mode network is that a lot of what it does, when you are not doing anything in particular, is practice social interaction, because that skill is at the core of human thriving.

So the order, lately, is relationships first — strong ones, where you sacrifice nontrivial time, attention, and resources on behalf of other people. He does not think we put nearly enough into that, and he worries that social media is thinning it. On that foundation, you ask what to do with the surplus: build a fire, build some tools. Everything else in a productivity system assumes the average case, that things keep going roughly as they have been. Put relationships first and you are doing the worst-case analysis. Boredom, impact, output, those drives are real and also not nearly as clear-cut. “We need people.” Fridman asks whether he is afraid of death. Newport’s move is to make it a productivity hack: confront it now, live and work in a way you can be proud of, so that when the confrontation is no longer optional you are more likely to feel all right about the situation. He cites David Brooks on eulogy virtues versus résumé virtues. Living by a code, even when it is hard, produces a kind of resilience religion used to structure and that soldiers still talk about. In its absence you need a replacement.

Fridman pushes on the video-game smell of all this: high scores, principles, an implied immortality behind every hack. Newport does not claim an answer. What he will defend are intimations — this feels aligned, that feels wrong, courage in a particular moment feels like contact with something. Those intimations are a grounding against arbitrariness. His grandfather was a theologian, and he recommends Karen Armstrong’s *The Case for God*. Her argument, as he tells it, is that to understand religion you have to go back before the Enlightenment, when it was an operating system for making sense of intimations, not a set of evidential propositions. Ritual, belief, and action were tools for living in alignment with what could not be said. Sam Harris’s critique, in that frame, misfires because it treats religion as an assent to unevidenced facts, which is an Enlightenment demand and not the original context. We are, Newport suspects, still searching for a good “live as if X is true.” Maybe artificial intelligence becomes the new god we were looking for. Maybe it just spits out 42. Fridman, citing Douglas Adams, had voted for 27.