Tech • AI • Robotics
Computer History in ChatGPT
6/10TL;DR
Computer History on the macOS desktop app turns local desktop activity into searchable memories and a timeline that ChatGPT and Codex can use to answer questions, continue tasks and suggest automations.
KEY POINTS
How the feature works
Computer History can be enabled in the desktop app settings on macOS. It converts ongoing desktop activity into memories and a timeline that an agent can reference later, allowing follow-up questions and task requests to draw on recent work without requiring the user to restate context.
Context-aware task support
The feature can identify recent documents and conversations from local activity history. In one example, it pinpointed the last Google Doc being edited as VideoScript Computer History, then used earlier messaging context to confirm the file had been shared with Josh and Priya in a group direct message and updated permissions for viewer access.
Automatic activity recap
Computer History can also generate concise summaries of how time was spent. A sample morning recap highlighted work on a Computer History launch video, including script drafting and revisions with teammates, plus prototyping a 3D World Builder in Codex and triaging feedback about usage limits.
What it captures
The system does not depend on screen recordings or audio recordings. Instead, it logs interaction events such as clicking, typing and app switching, an approach presented as faster and more efficient for building memories while preserving a more structured activity record.
User controls and exclusions
Users can block access to specific websites or entire apps. Those exclusions can be managed either from a new menu bar entry for quick changes or from the desktop app settings for more explicit control over what Computer History can and cannot reference.
Review and deletion tools
The settings area includes a history section where captured activity can be reviewed. Users can delete individual entries, clear recent history in bulk, or inspect the underlying memory files stored on the local file system to see exactly what has been recorded.
Automation and skill suggestions
Based on repeated patterns of work, Computer History can suggest skills and automations. Proposed uses include generating a recurring morning standup draft for a team or creating reusable skills from recently completed workflows.
Personalization for tool use
The feature is designed to help the agent adapt to a user’s habits, including preferred drafting apps and common document workflows. That added context is meant to make interactions more natural by letting the agent tailor its tool use to established routines instead of starting from scratch each time.
CONCLUSION
Computer History expands ChatGPT and Codex from prompt-driven assistants into tools that can work with ongoing local context. The central question will be whether its convenience, transparency and local controls are enough to reassure users about continuous desktop activity logging.
Explain this