One Screen Is Enough: An Attention Architecture for Mac

Does a second monitor fix multitasking? Often not — more screens, more windows, more scattered attention. What needs designing isn’t the physical layout of screens but the information architecture of your attention.

Start with the price of switching

Cognitive science is consistent here: the cost of task switching isn’t the act — it’s reloading the context. One interruption while coding costs minutes of lost reasoning. Which means:

• ⌘Tab to check if the AI job finished: three seconds of motion, plus several minutes of re-focus;
• Thirty switches a day cost you one to two hours of high-quality focus;
• But looking without switching costs essentially nothing.

So the goal is precise: separate “need to know” from “need to act”, make the first non-interrupting and the second low-friction.

A three-layer architecture

Sort every window by how you need it:

Layer 1 — The main task (owns the screen) The one thing you’re deep in: code, a document, a design. It deserves fullscreen and exclusivity. Rule of thumb: if you’ll type into it continuously for more than 15 minutes, it belongs here.

Layer 2 — Status lights (mini windows, pinned on top) Windows you need to sense but not constantly operate: AI job progress, CI pipelines, a match, market feeds, chat channels. Their proper form is a mini window — small enough for peripheral vision, floating above Layer 1. Mentions and completions register with a glance.

Layer 3 — Staging (close freely, recreate in one second) The “I might need this later” windows. They shouldn’t occupy any screen — but closing them shouldn’t cost anything either. That’s what a hotkey is for: ⌘⌥P miniaturizes the current window in a second; open what you need, close what you don’t.

Why PiP is the missing part of this architecture

Layer 2 has always been the hard one. Before window-level PiP, a Mac gave you two options:

• Switch to look (breaks attention);
• Split the screen (taxes the main task, and collapses past two windows).

A mini window is the only shape that is both non-interrupting and space-free. That’s why we built PipNow: turn any window into a floating mini window — hover for detail, click to interact — and Layer 2 becomes something you can actually keep around.

Day-to-day: a checklist

• Exactly one fullscreen main task at a time. That’s the discipline;
• Keep mini windows to three or fewer — beyond that, “peripheral” becomes “visual noise”;
• Ask of every mini window: “Do I need to know this, or am I just reluctant to close it?” Reluctant ones get closed;
• During deep work, allow yourself to close all mini windows — recreation costs one second; deliberately unload that guilt;
• Attention has an internal limit: however large the screen, you process one task at a time. Refusing to accept that, extra hardware only adds chaos.

Appendix: three-layer setups for common scenarios

Scenario	Main task	Status lights (mini windows)	Staging
Deep coding	Editor, fullscreen	AI agent, CI log	Docs, chat
Online course	Practice environment	Course video	Reference pages
Trading	Terminal	Feeds, news, chat	Research PDFs
Writing	The document	Reference page, ambience stream	Asset library

---

This post is the method; the mechanics live in the practical tips. For why this shape was missing until now, read why we built PipNow. And if you’re a split-screen user, this comparison explains which layer belongs to which tool.