OpenAI's Computer History maps desktop context

Computer History could make desktop agents dramatically more useful, but its success depends on earning trust around highly sensitive behavioral data.

• –Persistent context can reduce repetitive prompting and help agents resume interrupted projects
• –Searchable activity timelines extend AI memory beyond chats, repositories, and uploaded files
• –Suggested skills and automations could turn observed workflows into reusable developer tools
• –App exclusions, deletion controls, and inspectable files are essential safeguards, not optional polish
• –Capturing typing and app switches creates serious privacy and security risks if access boundaries are unclear