
What happened
On Day 14 of OpenAI's '28 days of shipping', the Codex desktop app gained the ability to edit previously sent messages within a thread, per the official tracker note.
Why it matters
Users can fix typos or unclear instructions in long coding sessions, but how far edits affect past work is unconfirmed and may require manual checks.
WHO IT HITSDevelopers and coding teams using the Codex desktop app for long tasks will be able to correct earlier instructions without starting over, though they may need to verify that prior work reflects the changes.
Summaries like this, every morning.
Day 14's update is a usability change for Codex desktop, not a new model or reset. It follows a series of daily announcements under OpenAI's '28 days of shipping' banner, which have included model speed improvements, context window expansions, and global resets on Days 6, 8, 10, and 13. The editing feature is distinct from those resets and from existing conversation controls like Steer, Queue, and Fork. Steer lets users redirect an in-progress task, Queue schedules a follow-up instruction, and Fork branches a conversation from an earlier point. Editing a past message, by contrast, changes the original instruction itself. Editing a message does not automatically undo any of those side effects, so users are advised to check Git status and GitHub separately; Codex can modify files, run tests, and interact with GitHub. Because OpenAI has not detailed how edits affect re-execution or history branching, testing the feature in a small, non-destructive thread before using it on production repositories is recommended.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Anthropic says the stretch Claude Code keeps working on one instruction grew 3.3-fold from March to September…

Nishika began offering an "AI agent" feature on SecureMemoCloud on October 13, 2026, available on all plans at…

HIX AI released Workspace 2.0, adding project-based task management, three agent modes (Lite, Full, Max), a li…

Oracle unveiled Oracle AI Workspace, which plugs OpenAI models and ChatGPT into Oracle Fusion Applications, Ne…

At the 先進的AI利活用アドバイザリーボード meeting on 2026年9月29日, officials presented a draft of items to flesh out guidelines…

Anthropic acquired Mithril Medicine, run solely by founder Cody Traumberger with Claude handling contracts and…
