OpenAI is adding a new layer of memory to ChatGPT’s desktop app on macOS. The feature, called Computer History, uses activity from your computer to help ChatGPT understand what you have been doing and what you may want to do next.
The idea is straightforward: instead of relying only on what you type into a chat, ChatGPT can refer to a timeline of your recent actions. That timeline can help it learn how you work, suggest automations, and pick up tasks that were left unfinished.
What Computer History Records
Computer History turns your actions into training data. According to the source article, that includes tracking clicks and keystrokes through the macOS desktop app, then converting activity into a timeline that ChatGPT and Codex can use when you make a request.
That changes the role of ChatGPT on the desktop. It is no longer just responding to prompts in isolation. With Computer History enabled, it can draw on a record of what happened across your recent work session and use that context to answer more directly.
The feature is designed to support practical requests. In a demo video, Dominik Kundel, a member of the Developer Experiences team at OpenAI, shows the app finding the last document he edited, checking whether it had been shared with people via Slack, and summarizing how he spent his morning.
Why The Timeline Matters
A timeline gives ChatGPT and Codex a way to connect scattered work activity. If a user switches between a document, a messaging app, and a coding task, the assistant can have more context about what was happening before the next prompt arrives.
That context could make follow-up requests less repetitive. Instead of asking the user to restate what they edited, where they shared it, or what they were working on earlier, ChatGPT can refer to Computer History when answering.
The source article describes three main uses for this context:
- Learning how you work based on your activity.
- Suggesting automations that fit those patterns.
- Picking up tasks that were left half done.
Those uses also explain why the feature is sensitive. Activity history can be useful because it is close to the user’s real workflow. That same closeness is why controls and boundaries matter.
Privacy Controls Are Built Around Opt-In
Computer History is opt-in, rather than opt-out. That means users must choose to turn it on before the feature starts building this activity timeline.
The source article also says users can exclude certain apps and websites from Computer History. That gives broader control over what kinds of work activity are available to ChatGPT and Codex.
For more specific control, users can delete entries. That matters because not every privacy decision fits neatly at the app or website level. A user may want some activity from a tool to remain useful while removing particular items from the record.
Ari Weinstein, Product and Engineering manager at OpenAI, said on X that Computer History will automatically ignore content in incognito or private browser tabs. That adds another boundary around browser activity, at least for tabs opened in those private modes.
How It Compares With Windows Recall
The feature invites comparison with Windows Recall because both involve an AI system using computer activity as context. The source article makes that comparison directly, while also drawing a key distinction in how the data is captured.
Microsoft’s controversial AI feature relied heavily on screenshots. OpenAI says Computer History does not capture images, videos, or audio. Instead, it relies on “events.”
That difference is important, but it does not make the privacy questions disappear. A system that understands clicks, keystrokes, app activity, and workflow events can still hold a revealing picture of how someone works.
For users, the practical question is not only whether Computer History captures screenshots. It is whether the timeline is useful enough to justify turning it on, and whether the available exclusions and deletion controls match the way they separate personal, private, and professional work.
What Users Should Watch
Computer History points toward a more context-aware version of ChatGPT on the desktop. The assistant can become more useful when it knows what task came before the prompt, which document was recently edited, or what work was left unresolved.
At the same time, this is a feature built on activity tracking. The opt-in model, app and website exclusions, entry deletion, and private browsing handling are central to how users will judge it.
The core tradeoff is clear from the source: Computer History can make ChatGPT and Codex more aware of your workflow, but it does so by turning computer activity into data the assistant can reference. Anyone considering it should understand both parts before enabling it.