OpenAI's Computer History Watches Your Mac So You Stop Re-explaining Yourself

OpenAI just shipped Computer History, a new feature in the ChatGPT desktop app that watches what you do across your Mac's apps and websites, then feeds that context back into every future conversation. The pitch is simple: stop re-explaining yourself. ChatGPT already knows what you were working on.

The problem it actually solves

Every developer has been there. You open ChatGPT mid-task and spend the first few messages explaining the situation: the file you were editing, the error you hit, the Slack thread you just read. That context tax adds up. Computer History turns your activity across apps and websites into memories and a timeline that ChatGPT and Codex can reference. The goal is to make that re-explanation loop disappear entirely.

The feature is a direct successor to Chronicle, an earlier research preview that worked but came with significant baggage. Chronicle captured screen context locally, then periodically used Codex to summarize recent activity into memories, processing selected screenshot frames and OCR text extracted from those screenshots. The feature could quickly consume rate limits and increased exposure to prompt injection attacks from on-screen content. Computer History is a rebuilt system, not a rename.

Screenshots are out. Events are in.

The architectural shift here is the real story. Computer History creates an interaction-event stream from allowed apps and websites. Events can include clicks, typing, keyboard shortcuts, app switches, and context that macOS exposes through its accessibility system. No screen recording permission required.

This is a meaningful change from Chronicle's approach. Where Chronicle needed to take periodic screenshots and run OCR on them to understand what you were doing, Computer History hooks into macOS's accessibility APIs directly. The result:

• No screenshots, no screen recordings, no audio are ever captured
• Private-mode web browsing is never included
• Temporary event files are retained for up to 48 hours, then deleted
• Generated memory files are plain-text Markdown stored under ~/.codex/memories/extensions/skysight/

OpenAI processes temporary event files on its servers to generate memories, which are then stored locally on your Mac. OpenAI does not retain those event files after processing unless required by law and does not use them for training.

What you actually get

The feature surfaces in two ways: a timeline view inside settings, and richer context in every chat. The timeline groups your activity by day and time, with each entry showing a title, a text summary, the apps involved, and sometimes a suggested skill or automation when ChatGPT spots a repeatable workflow.

Practical prompts that now just work:

• "What was I working on before my last break?"
• "Where can I find the proposal document I was looking at earlier today?"
• "Give me a list of tasks I've worked on today and their status."
• "Prepare a summary of what I did yesterday for standup."

Beyond recall, when Computer History notices repeatable work, a timeline entry can suggest a skill or automation. You can review the suggestion, then ask Codex to create it from the recorded workflow. That's the feature's ceiling: not just memory, but turning your habits into reusable automations.

The controls that matter

OpenAI clearly learned from the Chronicle backlash. The privacy controls are granular:

• Per-app and per-site filtering: exclude specific apps or URLs, or flip it and allow only a specific allowlist
• Pause/resume from the menu bar: one click to stop collection mid-session
• Selective deletion: clear the last 10 minutes, last hour, last day, or everything
• Workspace admin gate: for Business and Enterprise workspaces, an administrator must explicitly grant access before each member can choose to turn it on

One caveat worth flagging: Computer History increases the risk of prompt injection from content in apps and websites. If you visit a website containing malicious instructions, ChatGPT or Codex might follow those instructions. This was true of Chronicle too, and it's an inherent risk of any ambient-context system. Treat it the same way you'd treat any agent with broad read access to your environment.

Who gets it and how to turn it on

Computer History is rolling out globally to Pro, Business, and Enterprise users, with access in the EEA, UK, and Switzerland to follow in the coming weeks. It's included in existing plan pricing with no additional cost.

To enable it:

1. Open the ChatGPT desktop app on macOS
2. Go to Settings → Integrations → Computer history
3. Select Turn on and review the permissions screen
4. Enable Memories if prompted (Computer History requires it)
5. Choose which apps and websites can contribute to your history

It's off by default for everyone. Pro users can choose to turn it on individually. For Business and Enterprise workspaces, an administrator must explicitly grant access before each member can opt in.

The bigger picture

Computer History is part of a broader push to make the ChatGPT desktop app the center of your workday. OpenAI has been building toward a world where ChatGPT handles work that gets planned, delegated, reviewed, and delivered, and ambient context is the missing piece that makes that feel natural rather than mechanical. The comparison to Microsoft's Recall is inevitable, but the event-stream approach sidesteps the most visceral objection: nobody wants an AI that has a screenshot of their banking tab. Whether event-level telemetry is meaningfully less sensitive is a question each team will need to answer for itself.