ChatGPT Can Remember Cross-Application and Website Activities
OpenAI has launched the Computer History feature, allowing ChatGPT to remember user activities across applications and websites on their computers.
This feature is an improvement based on the previous preview of the Chronicle research, reducing token consumption and enhancing privacy controls.
The new timeline view allows users to review their work records and create skills based on frequently performed tasks.
Users can clear all or part of their history, choose to include or exclude specific applications and websites, and can pause or resume recording at any time.
The feature captures activities through macOS accessibility events instead of screenshots, is turned off by default, and is available to Pro, Business, and Enterprise users.
In market mechanisms, desktop AI assistants enhance personalized efficiency through local activity memory, using subscription flows with strong privacy controls; the beneficiaries are professional users seeking workflow automation, while traditional AI tools relying on manual context face pressure.
Source: Public Information
ABAB AI Insight
OpenAI previously launched Chronicle, which relied on screenshots to generate memory, but faced limitations due to privacy and cost issues. Computer History shifts to event stream capture and strengthens user control, continuing the iteration from experimental screen perception to controllable local memory.
In terms of capital and product path, the company aims to broaden the feature's reach to more paid and enterprise users by reducing token consumption and providing granular privacy options, motivated by enhancing long-term stickiness and the value of skill automation, strategically transforming desktop activities into a reusable personal knowledge base.
Similar cases can be seen in the controversies and adjustments of Microsoft Recall, as well as the evolution of various local memory tools; current AI assistants are in a phase of expanding from conversational memory to full workflow context, with privacy control becoming a key factor for implementation.
The structural judgment indicates a technological replacement: shifting from screenshot OCR to event stream capture reduces privacy risks and computational overhead; the mechanism is that when AI needs to continuously understand user behavior, a locally controllable timeline becomes a bridge connecting short-term interactions and long-term skills.
ABAB News · Cognitive Laws
AI that remembers user behavior will ultimately change the way we work.
The more granular the privacy controls, the higher the acceptance of the feature.
A local timeline is the true entry point for personalization.