April 28, 2026
Your cursor stays, the comments don’t
Show HN: Drive any macOS app in the background without stealing the cursor
This Mac trick wowed nerds, sparked privacy side-eye, and got everyone asking what it’s even for
TLDR: A new tool says it can control Mac apps behind the scenes without taking over your mouse, which could make automation and AI helpers much easier to use. Commenters were impressed, but the thread quickly split over privacy concerns, big future predictions, and basic questions about why ordinary people would want this.
A new Show HN project is turning heads because it claims to control macOS apps in the background without hijacking your mouse — basically, software that can click around for you while you keep using your computer like nothing happened. The makers are pitching a full toolkit for building and testing computer-using AI agents across Mac, Windows, Linux, and Android. But in classic internet fashion, the real fireworks are in the comments, where admiration, suspicion, and confusion all showed up to the party.
The biggest applause came from people who immediately saw the power move: running lots of app tests at once without your cursor getting yanked around like it’s possessed. One self-described ex-Apple engineer gave it serious respect and said that parallel app testing was the dream — but then came the sharp elbow: telemetry on by default. That kicked off the familiar privacy mini-drama, with the vibe being, “Cool tool, now stop phoning home unless I say so.”
Others went full hype, calling it one of the coolest hacks they’ve seen lately, while one commenter basically framed it as a warning shot to Apple: if Apple won’t make room for AI helpers, maybe Linux or Android will steal the spotlight. And then there was the wonderfully chaotic reality check from the peanut gallery: wait, what is this actually for? One person flat-out asked what makes it “for agents” instead of just normal automation, while another newbie begged for real-life examples. So yes, people are impressed — but they also want receipts, use cases, and fewer sneaky defaults.
Key Points
- •The article presents Cua as a cross-platform framework for building and deploying computer-use agents across cloud and local sandboxes.
- •Cua's API supports programmatic interaction with environments through shell commands, screenshots, mouse input, keyboard input, and mobile gestures.
- •CuaBot provides a CLI for running agents and GUI workflows inside sandboxes with native desktop windows, H.265, clipboard sharing, and audio.
- •Cua-Bench supports evaluation of computer-use agents on benchmarks such as OSWorld, ScreenSpot, and Windows Arena, and can export trajectories for training.
- •Lume manages macOS and Linux virtual machines on Apple Silicon using Apple's Virtualization.Framework, and the project is distributed under MIT with some third-party licensed components.