Tuesday, 8:05. You read the briefing, it took four minutes, and you closed it. The next one comes tomorrow morning. Between now and then, mail keeps arriving, and you’ve agreed to leave it alone.
That agreement is the whole return on the arrangement. It’s also the part that breaks first. The chapter before this one ended on the hole it leaves. The message that defines your Tuesday lands at 11:15, and it was absent from the briefing for a good reason: at 7:40 it didn’t exist yet. What you do about that is a list of exceptions, and the list has a wrong answer on both sides. Too many, and you’ve moved your inbox onto a page and kept all of its interruptions. Too few, and the first real emergency sits for eighteen hours, which teaches you permanently to check in between.
- The test is whether the next few hours would go differently if you knew now. At 8:05, that means differently from how they’d go if you found out tomorrow.
- The length of the gap sets the length of the list. A rule written for somebody who opens their mail four times a day fits that day, and borrowing it for a once-a-day read is the usual mistake.
- Every line has to be recognizable from outside the message. A sender, a date, a named condition. “Anything really urgent” is a feeling, and a feeling asks you to read the message first, at 11:15, while you’re sitting with a client.
- The two errors are different sizes. A false interruption costs you twenty minutes. A miss costs you the whole arrangement. So write the list generously, and cut it with two weeks of evidence rather than with nerve.
- Nearly every real emergency has a phone number attached. That, rather than discipline, is what makes a short list safe.
One neighbor is worth naming before you start. It answers a question that sounds identical and turns out to be a different one. If what you want to settle is how often to open your inbox at all, that’s when to check email, and it works from a schedule of passes rather than from a single read. This chapter assumes you’ve made that decision in its strongest form: once, in the morning, and that’s the whole of it.
The gap is what you are ranking against
The first attempt nearly everybody makes is a list of important things. That produces a list you can’t use. Importance is a property of a message, and what you need is a property of a moment.
Take one message and move it around the day. The controller writes to say a client’s depreciation schedule has changed. It matters, and it changes something.
At 8:05, five minutes after you closed the briefing, the honest question is whether anything you do before tomorrow morning turns on knowing it. Usually it doesn’t. You read it at 7:40 tomorrow and handle it then, and the outcome is the same.
At 8:50, on a day with a ten o’clock about those numbers, it’s the most valuable thing that could reach you. It changes what the meeting is for, and you have seventy minutes to do something about it.
At 4:50 in the afternoon it goes quiet again. Anything you’d do with it tonight keeps until 7:40, and knowing at 4:50 buys you an evening of thinking about it.
Same message, three verdicts, and all three turned on the same thing. The distance to the next moment you were going to find out anyway, and whether anything sat inside that distance.
So the working test is narrow. Would the hours before the next briefing look different if I knew now? Not whether it matters, and not even whether it’s urgent. Urgency is closer, and it’s still a fact about the message rather than about your day. Pulling importance and urgency apart is the input to this question rather than the answer to it. If the two still feel like one thing, important versus urgent separates them properly, and it’s worth reading before you write anything down.
Now the part that gets inverted. People adopt a briefing, feel the relief of it, and shorten their interrupt list, on the reasoning that the whole point was fewer interruptions. It runs the other way. A firm that reads its mail at 8:00, 11:00 and 2:30 has an exposure of about three hours. Whatever arrives inside one of those windows waits at most three hours to be seen. A firm that reads once at 7:40 lives with two gaps. The overnight one is mostly harmless, because the world is asleep in it. The daytime one runs roughly nine working hours, and that’s the one that bites. Nine hours is long enough to miss a same-day cutoff from a standing start. It’s long enough for a client to give up on you and call somebody else.
That’s the real cost of the arrangement, and it’s paid in exactly one place. The list. A once-a-day reader needs a longer exception list than a three-passes reader. A list of two lines, written in the first flush of a calm Tuesday, is how people end up abandoning the whole thing in week three.
Price the gap against your longest one rather than your average. For most people that’s a weekend or a travel day. Coming back from a real absence is a separate problem, and it has its own answer in the last chapter of this series.
What a national alert system had to settle
The clearest worked example of an interrupt threshold in American law sits a long way from email. It’s the rule that decides what may light up a phone lying face down on a table.
47 CFR 10.400 requires a participating carrier to “receive and transmit four classes of Alert Messages” (checked September 6, 2026):
- Presidential Alert
- Imminent Threat Alert
- Child Abduction Emergency/AMBER Alert
- Public Safety Message
Four classes, and the list ends there. Three properties of how it’s built are worth taking.
The set is closed. Every alert is one of the four, and something important that fits elsewhere stays off the phone. A list with an “other” line on it is a habit of promoting whatever is in front of you, and the promoting happens in the moment when your judgment is worst.
The threshold is a conjunction rather than a score. An Imminent Threat Alert “is an alert that meets a minimum value for each of three CAP elements: Urgency, Severity, and Certainty.” Urgency must be Immediate or Expected. Immediate means “responsive action should be taken immediately,” and Expected means “responsive action should be taken soon, within the next hour.” Severity must be Extreme, “an extraordinary threat to life or property,” or Severe, “a significant threat to life or property.” Certainty must be Observed, “determined to have occurred or to be ongoing,” or Likely, which “has a probability of greater than 50 percent” (checked September 6, 2026).
All three, every time. Severe and urgent still needs certainty. Urgent and certain still needs severity. Two strong axes sit below the bar while the third is weak.
Most people’s homemade lists use severity alone: the big client, the large sum, the name on the door. Urgency is the axis they add on the second pass, once something important and undated has interrupted a good afternoon. Certainty is the axis almost nobody writes down, and the list stands or falls on it. Certainty here means the thing can be identified as that thing from the outside, before anybody has read it. A rule you can evaluate only after reading the message leaves you with the reading, and the reading is what you came to save.
The bar for certainty is a coin flip. Greater than 50 percent, for alerts about extraordinary threats to life. That’s a startlingly low confidence threshold to find in a federal rule, and it’s the right one, because the two errors are different sizes. An alert that turns out to have been spare costs a few million people thirty seconds. An alert withheld because somebody wanted to be sure costs something permanent. Your version of that asymmetry is smaller and points the same way. A wrong interruption costs you a recovery, and a wrong silence costs you the arrangement itself.
One last detail sits at 47 CFR 10.280. Providers “may provide their subscribers with the option to opt out of the ‘Child Abduction Emergency/AMBER Alert,’ ‘Imminent Threat Alert’ and ‘Public Safety Message’ classes of Alert Messages” (checked September 6, 2026). Three of the four, and the Presidential Alert holds for everyone. Even a system this carefully bounded keeps exactly one line the subscriber simply lives with, and knowing which of your own lines that is turns out to be a clarifying thing to have decided in advance.
This rule governs commercial mobile service providers and public emergencies, and it stops there. The alerts concern tornadoes and abducted children. Your list is a smaller thing, and the stakes on it are your own. What transfers is the shape: a closed set, three tests that all have to pass, and a confidence bar set deliberately low because of which way the errors fall.
Protecting the window instead of judging the message
There’s a second way to build the rule, and it’s easier to hold in your head, because it leaves the ranking out altogether.
Rather than naming what is important enough to come through, name the hours when almost nothing is.
14 CFR 121.542 is the aviation version, and it’s short. “No certificate holder shall require, nor may any flight crewmember perform, any duties during a critical phase of flight except those duties required for the safe operation of the aircraft.”
Then it names, on the record, the specific things it rules out. The list reads: “company required calls made for such nonsafety related purposes as ordering galley supplies and confirming passenger connections, announcements made to passengers promoting the air carrier or pointing out sights of interest, and filling out company payroll and related records.” The next paragraph does the same for conversation, ruling out “eating meals, engaging in nonessential conversations within the cockpit and nonessential communications between the cabin and cockpit crews, and reading publications not related to the proper conduct of the flight.”
Then it defines the window mechanically. Critical phases of flight are “all ground operations involving taxi, takeoff and landing, and all other flight operations conducted below 10,000 feet, except cruise flight” (checked September 6, 2026).
Two design choices there, and both are borrowable.
The window is defined by a measurement rather than by judgment. The crew reads the phase off the altimeter, and the phase answers the question. Your equivalent is a statement about hours. Between 9:00 and 11:30 the bar is high, and at 4:30 it’s low. What changes across those two hours is what an interruption costs. The cleared block is the thing most likely to be eaten, and an empty afternoon is not a protected one covers what needs to stand in front of it before the hours actually hold.
The exception is stated as a function, and then the things that fail it are enumerated. “Required for the safe operation of the aircraft” is the whole test, and it’s one clause. Everything else in the section lists the things that most often talk their way past it. Look at what’s on that list. Galley supplies. Confirming passenger connections. Payroll. These are real work, with real consequences if they go undone, and every one of them was presumably being done at four hundred feet by somebody who thought it was fine at the time.
That’s the half of the writing people skip. A list of what may interrupt you is easy and mostly obvious. The part that holds up on a bad Thursday is the other list: the four things that most reliably talk their way through, named specifically enough that you recognize one on sight.
Take the analogy that far and stop. Section 121.542 governs certificate holders and flight crews, the phases are defined in feet, and the rule exists because of accidents. It’s a well-tested piece of drafting about aircraft, and the shape is the only part that travels.
A line you can apply without reading the message
Here’s the format, and the format is most of the work.
Each line needs three parts: the condition, how you recognize it from the outside, and what happens when it fires. The middle part is the one that decides whether you have a rule.
A line that fails: “Anything genuinely urgent from a client.” Applying that means reading everything, which is the state you came here to leave.
Lines that work look like this.
A named sender inside a named window. The e-filing acknowledgment address, during a filing week. The sender is a string, the window is a date range, and both read true or false on their own.
A named matter. Anything that mentions the Hallett closing, until it closes. Specific, self-expiring, and off the list in three weeks on its own.
A state you’re already tracking. Anything on a thread where somebody else is stopped until you answer. Blocked is a fact about the thread rather than a mood. If your system already records who’s waiting on whom, the rule reads that record rather than the prose.
A clock inside the current business day. Anything carrying a stated deadline that falls before this evening. A deadline you set yourself is one you can move, so it stays off.
The test for any candidate line is whether somebody else in the firm could apply it, from the outside of a message, on their own. A line that needs you is a preference rather than a rule, and it gets applied inconsistently. That’s the worst of the available outcomes: you get the interruptions, and you go on checking anyway.
Which categories deserve a line in the first place is worked through in ten correct alerts, and the morning they cost you, along with the harder half of what stays deliberately off. That piece also takes up who in a firm is allowed to add to the list at all. That last one matters more than it sounds. In a firm of two or four, a list anybody may extend covers a good share of ordinary traffic by the end of a quarter, one entirely reasonable addition at a time.
One thing stays off regardless of how you write the rest: the sender’s own opinion. A high-importance flag, or the word URGENT in a subject line, is a claim made by a party with an interest in the answer. Two of your forty clients use it the way it was meant. The rest use it for everything, or for nothing. And the ones who genuinely need you inside the hour are already on the phone, for the reason in the next section.
Once a line is written, the thing it becomes in software is a standing watch. A condition described in plain words, sitting there, firing when it’s met. The bar a line clears before it becomes one sits in setting your first watch in plain words, along with why nine of them is already an admission of something.
The phone is why a short list is safe
The strongest argument for keeping this to a handful of lines is the phone. Emergencies travel by phone, and they always have.
A building floods. A filing is rejected an hour before the cutoff. A closing collapses. A client is served with something. In every one of those, the person on the other end reaches for a phone. Email is asynchronous by construction and everybody already knows it, which is why the genuinely frightening things arrive as a ringing handset.
So the phone is your safety net, and the exception list covers one narrow band. Things that will matter within hours, and that come by mail rather than by call, usually because the sender has no idea they’re urgent to you. The acknowledgment that stayed silent. The counterparty’s attorney attaching a revised schedule at 2:00 for a 4:00 signing, with no note. That band is real, and it’s small, and its smallness is the entire reason a short list is a defensible thing to live behind.
Relying on it brings two obligations, and both have to hold.
You have to actually answer the phone. Somebody who reads mail once a day and screens calls has closed both channels, and the people who deal with them work that out inside a month. What they do next is treat you as out of reach, and that costs you in ways that never show up as a missed message.
You have to say so out loud. Say it as a standing description of how you work: mail is read in the morning, and anything that needs you sooner goes to the phone. An away message reads as absence, and a standing description reads as a practice. A good share of the pull back toward the inbox is worry that somebody is waiting and has no idea when you’re coming, and a published rule converts that worry into a fact. A firm norm holds better than a personal habit here, which is the practical half of setting client response-time expectations.
Your phone already ships with a threshold of exactly this shape, and most people build something more elaborate on top of it. Apple’s iPhone User Guide describes a Focus setting that lets through “repeated calls (two or more calls from the same person within 3 minutes).” It also notes that “regardless of your Focus settings, you can allow calls from your emergency contacts to come through” (checked September 6, 2026). The first of those is a rule almost nobody wrote down, and most people live under it anyway. It’s a good one. A caller who tries twice inside three minutes has said something about urgency that a subject line can say no other way, and they’ve said it knowing nothing about your list.
Two weeks that tell you which lines to cut
Write the first version generously. Six or seven lines, including two you suspect are too broad. Then measure it. Adjusting it by feel is what produced the forty minutes this series started with.
For two weeks, every time something reaches you between briefings, write one line in a notebook: what it was, and what you did in the following hour. At the end, sort them into three piles.
You acted within the hour. The line earned its place. Leave it.
You acted, but the same item at the next briefing would have gone exactly the same way. Cut the line. This is almost always the biggest pile, and it’s the one people argue with, because acting on something feels like proof that hearing about it early was necessary. The question is whether the item deserved attention then.
You didn’t act at all. Cut the line, and look at why it was written. A category that fires and produces nothing is usually a worry rather than an obligation.
That second pile is the whole finding, and it’s why the audit lives in writing rather than in memory. What sticks is the one time the list saved you, and that’s exactly the sample that makes lists grow. The interruption you handled competently, and could have handled tomorrow, is the one you forget.
A near neighbor proposes a similar log for a different question. Ten correct alerts asks whether what’s being captured is right at all, by counting how many prompts are worth acting on in any timeframe. This audit assumes the capture is fine and asks only whether the threshold sits too low. If both piles look bad, do that one first. A threshold tuned on top of bad capture is a setting on a broken instrument. That piece also carries the evidence on what happens to a person’s willingness to attend to a channel once it has been overdrawn, which is the mechanism you’re working to avoid. It carries a second finding too: turning everything off entirely produces its own anxiety rather than calm. Silence sits at the far end of the dial, and it’s its own way of not knowing.
Then set a date to look at the list again, and put it on the calendar rather than in your intentions. A quarter is about right. Exception lists grow on their own, because every addition is made in a week where it seemed obviously necessary, and deleting a line on the day it might have mattered takes a calendar entry to force.
Where Point takes this decision
Point treats the threshold and the briefing as one setting rather than as two products. Where the bar sits is yours to place. The things below it hold until the next read, and that holding is what makes the quiet between reads mean something.
A line becomes a watch, described in your own words. Ask to hear when a particular email arrives or a condition is met, and Point tells you at the moment it happens rather than at the next read. The alert speaks once, about the matter rather than about each record that changed. That’s what keeps a single confirmation from reaching you four times under four different names.
Point is the mail client, so the two halves agree by construction. Whatever decides a message is worth interrupting you for is the same thing that ranks it in the briefing. One judgment, used twice. The messages that stayed below the bar sit in the same feed, one move away, so the silence is checkable. On the phone, the same rule holds. The screen stays dark except for the things you named, and one tap opens the item itself rather than a list you then work through.
Two settings sound related and are kept deliberately separate. Point’s reach is set for each kind of work on its own, and each setting has three stops: propose and halt, prepare and hold, or proceed. Every one of them begins at the middle stop, so a setting above it is one you raised yourself. Each thing Point did is recorded in sequence, and an undo starts from that entry. Mail already delivered to another company’s server is past recall, which is the honest exception. All of that sits apart from the interrupt threshold, which covers what reaches you and when. Raising the dial leaves the bar where it is, and lowering the bar leaves the dial where it is. The benefits page carries the rest of the inventory. The mailbox underneath stays the Gmail or Microsoft 365 one you already have, with everything in place and the address you use now.
Deciding which of your obligations is worth an interruption stays with you. Point holds the list, applies it consistently at 11:15 while you’re with a client, tells you once, and keeps the record of what it raised and what it held. The lines on the list are yours, and they should be. They’re a statement about your business rather than about your mail.
The interval every line assumes
Take the list you just wrote and change one number in the arrangement around it.
Everything above was priced against a single briefing at 7:40, and the next one at 7:40 the following morning. Now put a second briefing at 1:00. The overnight gap holds steady. The daytime gap halves, from about nine working hours to four and a half, and half the list becomes optional in an afternoon. Every line that existed because nine hours is long enough to miss a same-day cutoff can go. What survives is the short set of things that need you inside four hours, and that set is roughly what you’d have written on a napkin in the first place.
Which means the real decision sits upstream of the list. It’s the one you’ve probably made by default, and have been treating as fixed since the first chapter: how many times a day the day gets handed to you, and at what hours. Move that number and the exception list rewrites itself. Leave it at one and the list carries nine hours on its own.
Briefed on your own schedule is where that number gets chosen, and it’s the next chapter. Bring the list with you. It’s the thing the number gets tested against.