2 min read

ChatGPT now runs Mac apps and beat chess on test

PCWorld tested ChatGPT’s new desktop control on macOS. It played Chess and wrote in Journal, showing how far computer-use tools have come.

Image: PCWorld

OpenAI’s new ChatGPT super-app can now control desktop apps on macOS, and in PCWorld’s test it handled both Chess and Journal with little trouble.

Released last Thursday, the updated app includes the built-in Codex coding harness, ChatGPT Work, and a new computer use mode that can take over your mouse and interact with local apps. According to Ben Patterson at PCWorld, that feature feels markedly more capable than earlier attempts at desktop control.

To test it, Patterson ran ChatGPT on a Mac mini, switched to Work mode, selected GPT-5.6 Sol, and asked a simple question: “Can you play chess on my Mac?” After requesting permissions — including screen recording so it could see the desktop — ChatGPT opened the Chess app in under 30 seconds and prepared a new game.

Recommended reading

Kodak’s $100 cameras keep outselling Sony and Canon

ChatGPT chess
ChatGPT chess

PCWorld reports that GPT-5.6 Sol made eight moves in less than three minutes, with most of the delay coming from waiting for the Mac’s responses. It also identified the opening as the Winawer variation and adjusted its play accordingly.

That puts ChatGPT more directly alongside Claude, whose desktop app has offered similar functionality for months. Patterson says Claude had failed to move pieces when he tried the same idea four months ago, but a fresh test with Fable 5 went much better: Claude also played successfully and recognized the Italian Game.

Claude chess mac
Claude chess mac

The test went beyond games. Patterson next told ChatGPT to open Apple’s Journal app, create a new entry, and write about how it was feeling. It launched the app, started a fresh note, and wrote a short reflective passage beginning, “Today I feel curious and quietly energized …”

ChatGPT journal
ChatGPT journal

The result, PCWorld argues, is that computer use is now genuinely usable for some tasks, especially developer-oriented ones. OpenAI suggests uses such as testing desktop interfaces, reproducing bugs, and handling repetitive GUI-based work. Patterson is less enthusiastic about turning it loose on sensitive apps or personal data, especially given the permissions required to watch the screen and control software.

He came away impressed, but cautious: desktop agents may finally be practical, even if many users still won’t want them anywhere near apps like Notes or Quicken.

Tomas Berg

Computing Editor

Tomas lives in the terminal. He covers chips, laptops, and operating systems with a focus on performance and efficiency. He reads kernel changelogs the way other people read fiction, and he's always on the hunt for the perfect mechanical keyboard switch. If it processes data, Tomas has an opinion on it.

via PCWorld

// Keep reading