ChatGPT Gets Computer Activity Tracking: The Computer History Feature
Originally published in Turkish. Read the original →
OpenAI has added Computer History to the ChatGPT desktop app, a feature that tracks users' computer activity as events. Details of the new feature and its approach to privacy.
For a long time, AI assistants have been in our lives as reactive systems that respond only to the user's written commands (prompts). But to truly speed up users' daily workflows, AI needs to understand the context of the actions performed on the computer. In line with this vision, OpenAI has integrated a new feature it calls Computer History into the ChatGPT desktop app for macOS.
The new feature records user interactions such as mouse clicks, window switches and keyboard activity as structured "events", creating a dynamic activity timeline the AI can refer to.
What Is Computer History and How Does It Work?
Computer History, which OpenAI added to the ChatGPT desktop app, collects the user's operational steps on the computer in a time-indexed memory. Thanks to this structure, ChatGPT and OpenAI Codex models can use what the user is working on at the moment, or worked on in previous sessions, as live context.
- The system basically works on these principles:
- Event-Based Monitoring: The apps the user opens, the menus they click, the windows they focus on and file movements are indexed chronologically in text and event format.
- Context Memory: When a question is put to ChatGPT, the model considers not only the prompt but also the context created by recent activity when answering.
- Codex and Developer Integration: Unfinished processes in code writing and terminal commands, or technical documents that were researched, can be recalled directly by the developer assistant.
What Will It Bring to Daily Workflows?
Computer History focuses on three core scenarios aimed at increasing desktop productivity:
1. Remembering and Resuming Unfinished Tasks
On a busy workday, forgetting where you left off while switching between different projects, browser tabs and documents is common. With Computer History, it becomes possible to get clear answers instantly by asking ChatGPT questions like "What was I doing after the second table in the report I was just reviewing?" or "Which parameter did I change in the configuration file I edited yesterday?"2. Habit Learning and Automation Suggestions
Over time, the system learns the user's repetitive actions (for example, a routine of downloading certain spreadsheets every morning and converting them into a report format). In light of this data, ChatGPT proactively suggests custom automations, Python scripts or shortcuts to save the user time.3. Eliminating the Hassle of Explaining Context
In traditional use, when describing a problem to ChatGPT, you have to write at length about the background, the tools used and the steps before the error. With Computer History enabled, the AI already knows the recent activity on the system, so it can produce solution-focused answers directly.How It Differs from Windows Recall: No Screenshots, Only Event Logs
The Recall feature Microsoft announced for Windows 11 took screenshots of the user's screen at regular intervals and archived them after scanning them with optical character recognition (OCR). This approach raised serious cybersecurity and personal data privacy concerns.
OpenAI, on the other hand, takes a different architectural approach with Computer History:
| Comparison Criterion | Microsoft Windows Recall | OpenAI ChatGPT Computer History |
|---|---|---|
| Data Collection Method | Periodic Screenshots (Snapshots) | Event Logs (Semantic Activity Log) |
| Visual / Audio Recording | Yes (screenshots are stored) | No (no screenshots, audio or video are recorded) |
| Context Type | OCR-based visual text scanning | Structured API and system interactions |
| Agent / Model Integration | Windows Search / Copilot search | ChatGPT and Codex reasoning engine |
| Target Focus | Visually searching past moments | Workflow optimization and autonomous completion |
Privacy and User Controls
AI tracking computer activity naturally raises security and privacy questions. OpenAI offers users comprehensive control mechanisms here:
- Fully Optional (Opt-in): The feature is off by default; no activity is recorded without the user's explicit consent and approval.
- App and Website Filtering: Sensitive areas such as banking apps, password managers or private messaging software can be blocklisted one by one and excluded from tracking.
- Managing and Deleting History: Users can view the generated timeline whenever they want and delete records for specific time ranges or their entire activity history with one click.
Computer History, which OpenAI added to the macOS ChatGPT app, is a critical milestone in AI assistants evolving from a passive chat box into a proactive digital helper. Its focus on structured activity records instead of screenshots stands out as an important lesson learned from the privacy crises Microsoft Recall went through.
In the future, context memories like this will continue to form the core infrastructure that lets autonomous AI agents (Agentic AI) complete complex tasks end to end directly on operating systems.
