OpenAI says the ChatGPT desktop app for Linux is now in preview, folding ChatGPT, Work, and Codex into one desktop package for Ubuntu, Debian, and Fedora across x64 and ARM64, with .deb and .rpm downloads listed in the official announcement. The company pitches it as a workspace for projects, files, browser workflows, and running Codex alongside ChatGPT, which is a very efficient way to say, "yes, the app would like to know your computer better". Source receipts: the OpenAI community post and the Hacker News thread that promptly began stress-testing the social contract.

The delicious public drama is not whether Linux users wanted support. It is that support arrived carrying every cultural trigger at once. On Hacker News, commenters immediately sparred over Electron, memory use, sandboxing, packaging formats, and whether giving a model desktop access is productivity magic or a silk-screened invitation to regret. One commenter reported 2.2 GB of resident RAM across 11 processes before even sending a query. Another said the real point of the app is direct computer control. A third summed up the mood by calling it "spyware that drags your system down to the level of Windows". That is not a bug report. That is a genre.

Meanwhile, the official OpenAI thread reads like preview software doing preview software things. Users discussed Japanese and Korean Fcitx input issues under Wayland, with a workaround using --enable-wayland-ime, asked for Flatpak or AppImage packaging, and noted that CLI projects do not yet show up cleanly inside the desktop project's view. Publicly, then, the launch has two records: the polished one saying Linux is invited to the party, and the commentariat one asking why the party arrived as Chromium in a trench coat.

No allegations, just vibes grounded in source material: this is what happens when a company selling AI productivity ships a desktop app to the operating system with the strongest opinions about how software ought to behave. Linux got the app. Linux also kept the right to heckle.