Take two weeks off from the changelog and you come back to a different tool. The back half of July brought four releases worth knowing about, and I’ve grouped them here by what changes in your workflow rather than by version number.
Working across more than one repo
Version 26.727 of the ChatGPT desktop app shipped July 30 with multi-repo review. The changelog says you can “see all repositories in a multi-folder project and the lines changed in each one,” then hit Review to inspect diffs across those repos without switching between separate review views.
The same release added an Activity view in the sidebar. OpenAI describes it as a way to see which chats you engaged with recently and which ones need attention. That helps if you keep three or four threads running and lose track of which one holds the fix you need.
Two smaller items rode along. The built-in browser’s address bar now searches your history, or falls back to Google when nothing matches. Generated images open in an expanded viewer with Focused and Canvas views, and you can comment across images and send targeted edits without leaving the conversation.
Multi-repo review only works because of version 26.715 on July 23, which let a local project hold more than one folder. You pick a primary folder when you add them. Per the changelog, new chats, Git operations, and automatic discovery of AGENTS.md, skills, and config.toml all run against that primary folder, while secondary folders stay available for search, reading, and editing.
My read, not something OpenAI said: they’re building for people who never had “one repo, one thread” work in the first place.
Talking through work instead of typing
That same July 23 release put ChatGPT Voice in the desktop app. The changelog describes it as a way to “talk through work and coordinate tasks in Chat, Work, and Codex,” powered by GPT-Live. OpenAI says you can direct multiple agents running in ChatGPT Work or Codex with your voice, and that it listens and speaks at the same time.
OpenAI lists Voice as available on Plus, Pro, Business, Edu, and Enterprise plans in the desktop app, plus through Remote on iOS. If you’d rather describe a bug out loud while you’re still reading the stack trace, that’s the use case.
One sign-in across your stack
On July 29, OpenAI started rolling out Sign in with ChatGPT in beta. This one is broader than Codex: the changelog covers select plugins and partner sites, naming Airtable, GitLab, HubSpot, Notion, Supabase, and Vercel as the starting set. Connect a supported plugin from the plugin directory and you can “create or link an account with that service in fewer steps,” per OpenAI.
The permission scope is narrow. OpenAI says the partner receives only your name, email address, and profile picture when one is available, and you still review and approve each plugin’s requested access as a separate step. It’s a beta on a selective rollout, so your account may not have it yet.
If you’re on the CLI
Codex CLI 0.146.0 landed the same day. From the GitHub release notes:
- Name a new session with
/newor/clear, pin the threads that matter, and switch between side conversations without closing them. - Thread forking picks up paginated history, plus temporary forks that stay out of your thread listings.
- app-server can now connect to remote Code Mode hosts over WebSocket.
- Standalone web search works for compatible custom model providers.
- Codex can discover skills your executor provides and read their associated resources.
The fix list is long on two fronts you’ve probably been fighting. Proxies are now honored across authentication, plugin downloads, MCP authorization, remote execution, WebSockets, redirects, and LM Studio connections. MCP connection handling also improved, and the release notes say closed servers reconnect without restarting healthy connections.
Nothing here is a headline on its own, but stacked together these change how your next session is set up, especially if you’re juggling repos or living in the CLI. Full version history is on our releases page.