ChatGPT Linux Update Breaks Codex Tasks

// 1h agoPRODUCT UPDATE

ChatGPT Linux Update Breaks Codex Tasks

Multiple reports indicate Linux desktop version 26.924.22138 leaves Codex tasks stuck indefinitely, while downgrading to earlier builds restores functionality. OpenAI’s bundled Codex runtime appears functional, pointing to a desktop integration regression. [GitHub issue](https://github.com/openai/codex/issues/48417)

// ANALYSIS

This is a serious regression for Linux developers relying on local Codex workflows, especially because the app remains in preview.

–Reproduced across Fedora, Ubuntu, Debian-based systems, and multiple desktop environments
–Prompts can remain stuck at “Starting your task” with cancellation also failing
–Standalone Codex CLI reportedly works, narrowing the problem to desktop task dispatch or subprocess handling
–Downgrading to 26.917.71314 or 26.901.41600 is the practical workaround
–Linux support is still officially labeled public preview, so release stability remains a concern. [OpenAI release notes](https://help.openai.com/en/articles/6825453-chatgpt-release-notes)

// TAGS

chatgpt-desktopcodexai-codingcoding-agentdevtoollinux

DISCOVERED

1h ago

2026-09-28

PUBLISHED

1h ago

2026-09-28

RELEVANCE

8/ 10

AUTHOR

mark_k

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

UPDATE29m ago

Topview Plugin Puts Agents Inside Canvas

Topview Plugin connects AI agents to a Canvas workspace for planning, generating, editing, and organizing images, videos, ads, and campaigns. It supports batch ecommerce ads, reference-video remakes, and short dramas while keeping assets and versions connected.[Topview](https://www.topview.ai/plugin/chatgpt-plugin)

MODEL40m ago

Claude Sonnet 5.5 runs 30% faster

Anthropic launched Claude Sonnet 5.5, a faster, lower-cost complement to Opus 5.5 for coding, documents, spreadsheets, and design work. It runs over 30% faster than Sonnet 5 and can cost up to 30% less through reduced token usage.

TUTORIAL43m ago

CMU 11-768 Covers Full Agent Stack

Carnegie Mellon’s Fall 2026 AI Agents course, taught by Graham Neubig and Daniel Fried, is making lectures and course materials freely available online. The curriculum spans agent harnesses, tool use, evaluation, safety, and training models with SFT and reinforcement learning.