Blog · Product notes

Focus nudges and drift detection: how Modoi keeps you in deep work

The word "notification" is basically poisoned at this point. A decade of apps interrupting you to say something unimportant has trained most people to filter out alerts by default, which is the correct response. So when a productivity tool says it will "nudge" you toward focus, skepticism is reasonable.

Here is why focus nudges are different from what you are imagining, and why drift detection specifically is worth understanding.

What drift detection actually is

Drift detection watches what you are doing on your computer and compares it against what you declared you were going to do. If you said you are working on a design review and you have been in your email client for twelve minutes, that is drift. The system notices and tells you.

This is a fundamentally different mechanism from a Pomodoro timer, which counts down regardless of what you are doing. It is also different from Do Not Disturb mode, which blocks input but says nothing about your own behavior. Drift detection requires the tool to have an opinion about whether your current activity matches your declared task, and to tell you when it does not.

Modoi forms that opinion on your Mac, in two layers. A fast deterministic classifier watches the active window on every capture beat — it knows each category's lane (focus, neutral, or likely a distraction), tracks how long you have held the same page, and flags a mismatch the moment it becomes meaningful. On top of that, an on-device model checks in hourly to compare the last window of work against the task you declared. All of this happens on your Mac. None of it goes to a server.

Why on-device matters for drift detection specifically

Drift detection that relies on a cloud server has a real problem: your activity data has to leave your machine. For most professionals, that is a hard no. Even for those less concerned about privacy, the latency makes the feature less useful. A nudge that arrives a few seconds after you drifted does not feel responsive; it feels like an accusation after the fact.

Running detection locally means it is immediate and private. Modoi can observe your window context in real time without that data touching any external infrastructure. The hourly check runs against a model that ships inside the app — on a fresh Mac with no internet, the weights are already there.

There is also a practical reliability argument. If drift detection depends on an API, it stops working when you are offline. Modoi's focus features work without an internet connection.

What focus nudges look like in practice

Modoi does not bombard you. The design principle is minimum necessary interruption: one pause when the system reads meaningful drift, not a cascade of reminders every few minutes. The same page that just triggered the freeze is quiet for a while after you dismiss it.

The nudge appears as a screen freeze over the window you are on, with a small prompt: "Distracted?" and the app that tipped it over. It is easy to dismiss if you intentionally switched context — one click on "Thanks for the reminder!" lifts the freeze and lets the moment pass. If the app is genuinely not a distraction for you, "This is not a distraction" waves it through for the rest of the tracking session, so it stops counting against you.

That per-run learning is one half of the calibration. The other half is the tracking board: you decide which categories are focus, which are neutral, and which are likely a distraction, and both the drift signal and the screen freeze use exactly those lanes. Rather than applying one company's rulebook to everyone, Modoi applies rules you set for yourself.

How this compares to other focus tools

Most apps in this category work by blocking. Freedom, Cold Turkey, and similar tools prevent you from accessing certain sites or apps during a focus session. This is blunt but effective if your distraction patterns are site-specific and predictable.

Modoi's nudge is a moment of friction, not a lock. When drift reads, the screen freezes under a blur and one prompt asks whether the app you are on is a distraction. The freeze lifts the moment you answer — nothing is blocked, no site is locked, and you are never locked out of a window. The distinction matters if your distractions are not site-specific. Checking your task list obsessively, opening half-finished documents, switching between project contexts without completing anything: none of these are blockable because they look like legitimate work. Drift detection and your category lanes catch them anyway.

The tradeoff is that Modoi requires more setup than a blocklist. You need to arrange the tracking board — what is focus, what is neutral, what is likely a distraction — and Modoi is most useful when you keep that intention meaningful. If you want a tool that requires no intentionality on your end, a site blocker is simpler.

If you want a tool that works with how your actual workday is structured, drift detection is the better mechanism.

Frequently asked questions

Drift detection only works if you actually declare your tasks and take the nudges seriously. A tool that catches you drifting is not useful if you dismiss every alert and go back to Twitter. Modoi is most effective for people who already have some intention around their work and want honest feedback on whether they are following through.