OpenAI launches ChatGPT Work

OpenAI’s biggest developer-adjacent update this week is ChatGPT Work, a new agent layer inside ChatGPT aimed at multi-step execution rather than one-shot answers. The release blends Codex-style task handling with plugins, browser access, scheduled tasks, and desktop integrations so users can turn prompts into finished work products. For developers and technical teams, the notable angle is that OpenAI is pushing agent workflows beyond code generation into broader execution loops while still keeping Codex at the center of the product stack.

Key Features or Updates

ChatGPT Work can gather context from connected apps, browse the web, create artifacts like docs, slides, sheets, and sites, and continue working on longer projects in the background. OpenAI also says Codex is being merged into the new ChatGPT desktop experience, adding inline diff editing, PR review in the side panel, faster computer use, and multi-repo support.

Impact on Developers

This moves OpenAI’s agent story closer to a full operating layer for knowledge work and engineering-adjacent workflows, not just chat or code completion. Teams evaluating agent platforms will see this as OpenAI competing more directly on persistent execution, connected tools, and cross-app automation.

How to use it

Users can start by connecting plugins for sources like Slack, Teams, Drive, SharePoint, calendars, and internal tools, then give Work a multi-step request with approvals where needed. The desktop app is the most interesting surface for technical users because it combines local files, browser-based work, and Codex-style coding flows in one place.