Every version of this advice ends with "delete it." That works if the thing is optional. This page is for everyone it isn't optional for.
The problem
If your work lives on your phone, none of the standard advice survives contact with your Tuesday. The messages come in all day. The client calls. The photos are the proof you did the work. The invoice, the schedule, the code you need to log into the thing that pays you. Telling that person to delete their phone is telling them to delete their income, and they know it, so they discard the whole argument and keep scrolling.
The way it feels from the inside is being a whiskey taster who's trying to quit drinking. The thing that's hurting you is the same thing you get paid to hold. You can't put it down, so you conclude you can't fix it, and that conclusion is what makes it permanent.
The analogy is exactly right about the difficulty and exactly wrong about the outcome, and the place it breaks is the whole reason this page exists.
You can't separate the part of whiskey that pays a taster from the part that ruins them. It's the same molecule. But you can absolutely separate the part of a phone that runs your business from the part that eats your evening, because they were never the same thing to begin with. They just ship in one slab of glass.
So you don't have to quit the bar. You take the bottles out and leave the tools where they are. What follows is how.
The cut
Everything on the device is one or the other, and almost nothing in the way we talk about phones makes the distinction, which is why the conversation stays stuck.
A tool waits. You go to it with an intent, it does the thing, you leave. It never starts anything. It has no opinion about how long you stay. Your dialer is a tool. So is the camera, the map, the calendar, the banking app, the thing your POs live in.
A feed arrives. It serves you things you didn't ask for, it doesn't end, and its success is measured in how long it kept you. It isn't worse engineering than a tool. It's better engineering, aimed at a different target.
Two questions sort anything on your device in about four seconds:
Can it end? And did you ask for what it just showed you?
Anything that fails either one is a feed regardless of what it's called or what it's for. This matters more than it sounds, because the sorting doesn't respect categories. A work app with an activity stream in it's a feed wearing a hard hat. A messaging app is a tool right up until it has a discover tab. The label on the icon tells you nothing; the two questions tell you everything.
The build
Leverage order. The first step isn't a setting, it's a person, and without it every step under it's a ten-second undo.
The specifics below are iPhone. Android has equivalents under Digital Wellbeing and Family Link, but the names differ.
Start here. On an iPhone all of this lives under Screen Time, in Content and Privacy Restrictions, and the whole thing is held shut by a Screen Time passcode. If you're the one who knows that code, you haven't built a constraint. You've written yourself a suggestion, and you already know how those go at eleven at night.
So do it in one sitting with somebody else in the room. You make every change below while they sit there. When it's done, they type the passcode and they don't tell you what it is. A partner, a parent, a friend, anyone. They don't have to police you, they don't need to know why, and in all likelihood they will never be asked for it again.
This isn't a statement about your character and it isn't a trust exercise. It's the same reason somebody trying to stop drinking doesn't keep a bottle in the house. You aren't giving up the ability to decide. You're moving the decision out of the ten seconds where you're worst at making it, into daylight, where you'd decide differently and you know it.
If there's nobody, generate a long random code, write it on paper, and put the paper somewhere with real friction. A safe, a drawer at work, an envelope you mail to yourself. The requirement isn't secrecy. It's that getting it back takes longer than the urge lasts.
This is the step that does the most work. Without it every other removal is decorative, because a deleted app is one search away. In the same restrictions screen, the browser can be removed from the device entirely rather than hidden.
If your work needs the web, you don't have to give it up. The same settings let you allow a named list of sites and block everything else. Put your suppliers, your portals, your bank on it. That list is the whole difference between a device that reaches anything and a device that reaches your job.
But the allowed list isn't really where this is decided, because the browser is the one thing you can always justify. So it needs a test that runs in the moment rather than a setting, and it's two questions.
Almost nothing needs to be known right now. Curiosity isn't a bad thing and this isn't an argument against being curious, but curiosity in the middle of something else is the exact door the habit walks through, and it doesn't feel like the habit at the time. It feels like a legitimate question, which is what makes it effective.
Carry something to write on and put the question there instead. Then look the whole list up later, in one go, at a computer. Two things happen. The ones that mattered still matter and you answer them properly. Most of the rest you won't care about by the time you get there, and watching that happen a few times teaches you more about the habit than any screen time report ever will.
Otherwise every deletion below is a thirty-second undo, performed by a version of you who's tired and isn't going to argue the point. Restricting installs turns "delete the app" from a gesture into a decision with a cost attached.
Everything that survived the cut goes on a single page. No second page. If it isn't on page one, it doesn't exist. Then close the back doors, because they're the part people forget: the app library is a browsable list of everything you just hid, and search will launch anything by name whether or not it has an icon.
A notification from a person is a tool doing its job. A notification from an app is a feed reaching into your pocket to start something. Turn off everything that's not a named human addressing you directly.
This does more than it looks like. Unpredictable interruptions are the delivery mechanism for the variable-reward schedule that makes any of this sticky. Turning them off doesn't just reduce noise, it removes the reinforcement.
Nearly every serious work app now has an activity stream, a news tab, or a suggestions panel bolted on, because the people who build them read the same research. Turn those off where you can. Where you can't, go straight to the screen you need instead of landing on the home view and reading whatever it decided to show you.
Everything above removes places to go. This step is different: it works on the physical thing in your hand, and it's the part most people skip because it sounds cosmetic. It isn't. It's the cheapest friction available.
Grayscale. Colour is a real part of the pull. Bind it to the accessibility shortcut so you can flip it back for the thirty seconds when you need to see a photograph in colour.
A plain dark case. Nothing you enjoy looking at, nothing that makes it feel like an object worth holding. This is a hand tool now and it should look like one.
A matte screen protector. It kills the gloss, and it makes video noticeably less pleasant to watch, which isn't a side effect. It's the entire reason.
Raise to wake off. Tap to wake off. This is the one that surprises people. Both of those exist so the screen lights up before you've decided anything, and they're worth thinking about against the rule at the top of this page. A tool waits. Raise to wake means the phone doesn't wait. It greets you every single time you pick it up, and it has been doing that a hundred times a day.
With both off, the screen stays black until you deliberately press the side button. It sounds like nothing. In practice it's the single change you'll notice most, because it converts an ambient invitation into a small act you've to choose, and a surprising number of pickups die right there at the black screen.
What stays
The fear is that a device like this stops being able to run your work. Here's the whole ledger.
Read the second column again. There's nothing in it that you use to make money. That's not a coincidence and it isn't a trick of how I wrote the list. The feeds were never the part doing the work.
The easy ones
The phone is the hard case because you can't delete it. Nothing else in your house has that problem, which means the rest of this takes one evening and you should do it in the same week. Leaving one screen open is how people end up with a beautifully configured phone and an unchanged life.
Worth stopping on, because it's the strangest thing happening in media right now. Broadcast was the original bounded format. It had a schedule, it ended on its own, and when it ended you stood up and did something else. That ending wasn't a design choice, it was physics.
There's now an infinite feed of short video on the largest screen in most houses. The ladder didn't just add a rung at the top. It went back down and removed the stopping point from the one format that came with one.
So: delete the feed apps from the set. Not hidden, off it. Then in whatever you keep, turn off autoplay-next, which is the one setting that quietly converts an episode into a feed. If a service gives you no way to stop it from starting the next thing, that service is a feed and it can go with the others.
That's worth taking literally. Cancelling the all-you-can-watch services and buying individual episodes or films restores two things at once: a definite ending, and a small friction before each one. You choose the thing, you pay for the thing, you watch the thing, it stops. Most people find they watch a third as much and enjoy it considerably more, and the bill usually goes down rather than up, which surprises everybody.
Same diagnosis as the phone, easier treatment, because almost nobody needs a feed on a work machine. Block the feeds somewhere you can't casually undo, rather than relying on a browser extension you can switch off in two clicks. Log out of everything that's a feed and delete the saved passwords, so that getting back in costs a real minute instead of a keystroke. Install no feed apps at all; if something is worth reading on a computer, it's worth going to on purpose.
The test at the bottom of this page applies here too, and computers usually fail it worse than phones do.
The test
Not screen time. Screen time will drop and it's a lagging, noisy number that also drops when you've a busy week.
The test is this. Pick the device up with no intent at all. Don't go to anything. Just pick it up the way your hand does it a hundred times a day, and see where your thumb goes.
If there's nowhere to go, the build is right. You'll put it back down within a few seconds, faintly annoyed, and that annoyance is the whole point: the device has become a tool, which is a thing that's boring when you don't need it. Tools are supposed to be boring. That's what distinguishes them.
If your thumb finds something, you've found a feed you missed. Go sort it with the two questions and take it out. Then run the test again next week, because they come back. Apps add feeds in updates, and the pressure described on the front page doesn't stop just because you've finished configuring something.
Still being worked out
This page is the part I'm actively in the middle of rather than the part I've finished, and it will change as I find out which of it holds up. It's also the part where somebody else's experience is worth more than mine, because I've run one version of this on one device.
If you built one of these and something here's wrong, or a step didn't survive contact with your actual job, or you closed a hole I left open, I want to know. Corrections go straight into the page.
Everything that comes back and holds up gets folded in here, and if it changes my mind about something I will say so rather than quietly editing it.