screen time apps ask the wrong question

Every app that blocks distractions asks you the same question, in one of four accents.

The schedule asks which hours you would like to lose your phone. The quota asks how many minutes of a given app per day is acceptable. The session asks whether you are focusing right now, and waits for you to press start. The friction prompt asks whether you are sure, one app-open at a time.

They are good questions. Millions of people are better off for having answered them. But all four are versions of the same request: predict yourself. Tell the software in advance when you will be vulnerable, and it will stand guard at that hour.

That works when your vulnerable hour is predictable. Evenings in bed. The first twenty minutes of the workday. It works much less well for the moment I want to talk about, because that moment has no schedule at all.

the moment nobody can schedule

You give a coding agent a real task. It starts reading the repository, editing files, running the test suite. Somewhere between thirty seconds and five minutes will pass before it needs you again.

You did not plan this pause. You cannot predict the next one. It will happen again in four minutes, and maybe forty times before the day is over. Your hands are free, your attention is idle, and the phone is right there.

This is the gap that costs you, and it is worth being precise about why. It is not mainly the minutes; forty seconds is nothing. It is that the pause was not empty. You were holding a position: what the agent is doing, what you will ask it next, the constraint you remembered two prompts ago and never wrote down. A feed is extraordinarily good at overwriting exactly that.

why each approach misses it

Scheduled downtime is aimed at the clock, and the gap does not keep office hours. To catch it you would have to block nine to five wholesale — but you need your phone during those hours too, for two-factor codes, calls, the group chat that is actually work. A block you resent is a block you turn off by Thursday.

Daily app limits meter how long, never when. Thirty minutes of Instagram over lunch is a perfectly fine way to spend thirty minutes. The same thirty minutes taken forty seconds at a time, from inside forty separate gaps, is a completely different thing — and the limit counts them identically.

Manual focus sessions need you to press start at precisely the moment you are least likely to. You start a session when you already feel focused. Nobody opens a timer app, picks a duration and taps begin for a pause that might last ninety seconds; and anyone who did would spend the day starting and stopping timers instead of working.

Friction prompts get closer than the rest, and deserve the credit. They fire at the right moment, when your thumb is already moving. But the final step is still a question, asked of the one person in the building who has already decided they want to look. Sometimes you say no. Often you breathe, wait out the delay, and open it anyway.

a different question

The alternative is to stop asking you to forecast your own weakness, and instead ask the machine something it can actually answer: what is happening right now?

Your Mac already knows. A coding agent running is not a mood or an intention. It is a process, observable, with a beginning and an end. That makes it something software can react to without anyone having to predict anything.

So cantscroll blocks on that event instead. Claude Code, Cursor, Codex, Antigravity or Devin starts working on your Mac; the apps you picked lock on your iPhone; the session ends and they unlock on their own. You do not schedule it, start it, or decide anything in the moment — which matters, because the moment is exactly when your judgement is worst.

the blocking is the same. the trigger is the product.

Worth being clear, because comparison posts usually pretend otherwise: cantscroll does not block your apps better than Apple does. It blocks them using Apple’s Screen Time framework — the same machinery underneath most serious iOS blockers, including several of the apps in that table.

There is no clever new way to lock an app on an iPhone, and anyone claiming one is describing something Apple will eventually close. The entire difference is the trigger. Everything else in the category is triggered by a clock, a counter, or by you. This one is triggered by your laptop noticing that you started working.

where a normal blocker is the better tool

An honest comparison has to include the cases where the other side wins, and there are several.

If you doomscroll in bed at midnight, cantscroll does nothing for you — no agent is running, so nothing blocks. That is Downtime’s job, and Downtime is good at it. The two are complements rather than substitutes; I run both.

If your problem is total consumption rather than interrupted concentration, an app limit is the more direct instrument. And if you want distractions blocked while you write prose, or read, or do anything a process monitor cannot see, a manual session is the right shape and always will be.

cantscroll also needs both a Mac and an iPhone, paired, with a small free agent installed on the Mac. Today it shields apps on the phone; it does not block websites on your laptop. If your distraction lives in a browser tab six inches from the terminal, this is not yet the tool for that.

One more thing, deliberately: a pause button and a panic exit always work. It is a commitment device, not a parental control, and anyone determined to get past it will. That is a design decision rather than an oversight — the failure mode of an inescapable blocker is that you uninstall it in a moment of annoyance and lose the protection entirely.

what it comes down to

Screen time apps were built for a world where the distraction competed with your work. In agentic development it competes with your waiting — and your waiting, it turns out, is where a growing share of the thinking happens.

You cannot put that on a calendar. But your Mac can see it start.