OpenAI Codex’s Major Update: Screen‑Free Coding, AI “Goal” Mode, and PowerPoint Integration

Codex receives its biggest upgrade yet, introducing five new capabilities that dramatically raise its utility. The headline feature, Appshots , lets users press Command twice to let the AI read the entire screen—including off‑screen text, file paths, and URLs—and instantly generate corresponding code.

Appshots automatically attaches the captured context to the current Codex thread; if the user has interacted with that thread within the past 60 seconds, a new Appshot is appended instead of opening a new conversation. Multiple Appshots are also grouped together, and the feature works on any macOS version. By enabling the “Locked Use” option in macOS settings, Codex can operate on a locked or sleeping Mac, effectively allowing remote AI‑driven coding without unlocking the device.

The second major addition is the /goal mode , which graduates from experimental status to a production feature. Users set a clear milestone in the Codex app, IDE extension, or CLI, and the AI works continuously—day or night—toward that goal, providing progress updates, allowing direction changes, or pausing when needed. This transforms a single‑prompt interaction into a long‑running autonomous workflow.

On the same day, OpenAI launched the ChatGPT for PowerPoint plugin in global beta. A simple prompt can create new slides, edit existing ones, restructure content, and even polish language, with every AI‑generated element fully editable. The plugin can pull live data from Gmail, Outlook, and SharePoint, eliminating manual copy‑paste for quarterly reviews, board decks, or client presentations. This adds a third AI competitor—Claude and Copilot—to the PPT workflow market.

Usage metrics underscore the shift: weekly active Codex developers have risen to over 4 million, up from 3 million two weeks earlier, and the user base in ChatGPT Business/Enterprise grew six‑fold from January to April 2026. Notably, 50 % of Codex users now employ the tool for tasks beyond coding, such as automation, cross‑tool collaboration, long‑term task management, and remote computer control. Analytics have been expanded to track active users, credit consumption, token usage, run counts, code‑line generation, and plugin usage, with a new Analytics API for organizational insight.

Collectively, these updates illustrate Codex’s evolution from a “programming assistant” into a full‑stack AI work platform, laying a pathway from tool to teammate and, ultimately, toward capabilities that could outpace human productivity.