Wink Pings

He Turned the macOS Dock into a Mode Switch, with AI Deciding What You’re Doing

Thariq shows FocusDock, a personal experiment that uses an agent to customize writing and building modes, classify activity and notifications in real time, and switch window layouts and Chrome access scope. The community is already asking about compatibility and practical boundaries.

Thariq posted a personal experiment on X: FocusDock. It is a mode-based macOS Dock with presets such as writing and building. When you switch modes, the Dock doesn’t just swap out a set of icons. According to Nabeel Hyatt’s question and Thariq’s reply, it activates different window groups and layouts, mutes certain types of notifications, and restricts which websites Chrome can access. Thariq added that restricting Chrome may require an extension, and he’s still figuring out the details.

That detail says more about what the project is trying to do than the demo itself. FocusDock turns “I’m going to write now” and “I’m going to code now” into system-level executable states. Windows, notifications, and browser access all follow that state. Here, the Dock is not a launcher so much as a mode switch.

In the demo video, the terminal is in /Projects/focus-dock; running build-app.sh goes through 14 steps and completes the build in 8.41 seconds, generating FocusDock.app. At the bottom, you can see a row of app icons and the macOS logo. The video is from Thariq’s tweet.

In terms of implementation, Thariq says the agent can customize mode behavior and add new modes. It uses jev to classify his activity and notifications in real time, then executes actions. For ordinary users, you can think of it this way: the agent watches what you’re currently doing and decides which notifications to block, how windows should be arranged, and which sites the browser can open. For developers, modes are configuration, the agent is the interface for changing that configuration, and jev is the classifier. The really hard part is not generating the interface; it’s whether the classification is accurate and whether users can quickly correct it after a misjudgment.

The community response was straightforward. A DockDoor user asked whether it could support DockDoor after release, since they have highly customized settings they can’t live without. One user said their ADHD brain needs this. Addy Osmani said it looks good. Thariq replied that it works better than he expected. These responses point to the same need: reducing the cost of context switching.

Similar ideas are not new. Nate Parrott previously showed a concept that grouped tabs, files, and terminals by project and switched between them with a swipe. Someone in that tweet’s replies mentioned Slidepad, a browser tool that slides in from the edge of the screen. FocusDock’s difference is that it puts the agent into the Dock and the mode layer, letting AI help decide “what should be allowed in the current mode.”

The video is from Nate Parrott’s tweet.

Right now, FocusDock is still a personal experiment with no release plans. What’s worth watching is that it provides a concrete example: AI-native doesn’t necessarily start with a chat box; it can also grow out of old interfaces like the Dock. In the short term, it won’t replace the macOS Dock. What really needs validation is the accuracy of mode classification and whether users are willing to let an agent manage notifications and the browser. The Dock is an old place—so old that no one thinks it can be redone. Thariq’s experiment at least proves that one entry point for redoing it can be “what mode are you in right now.”

发布时间: 2026-10-28 20:42