Artificial intelligence technology company OpenAI has officially released the latest update for its desktop application Codex. The most notable change in this update is the official launch of a new feature called Appshots for the macOS version, designed to allow AI to directly read and understand user software application windows.
With this new feature, macOS users can quickly send a screenshot of the currently active application window to Codex by pressing the Command key on their keyboard. This move not only greatly simplifies the user's workflow but also marks a new level in the interaction experience of desktop AI.
Breaking Screen Limitations to Read Hidden Text
Notably, the Appshots feature does not just stop at traditional visual screenshots when transmitting data. While capturing the screen, the system can also deeply read the text content inside the window, including hidden parts that are temporarily not visible due to scrolling.
OpenAI stated that this feature was primarily designed to solve core pain points in users' real-world work scenarios. For example, developers can directly use this feature to transmit complex interfaces when debugging web pages in a browser or when designers need to implement complex interface designs, eliminating the tedious steps of manually taking screenshots and copying text.
Core Task Management Feature Now Official
In addition to the innovative window reading feature, the previously highly anticipated task management commands have also completed their experimental phase in this update and are now officially available to all users. Users can now set clear long-term goals within the Codex application, IDE extensions, or command-line interface.
Once the goal is set, Codex will continue to advance the task over several hours or even days until reaching the preset milestones. During this time, users can view the latest progress of the task at any time, adjust the subsequent execution direction, and even pause the task directly based on actual needs.
