Personal activity search and memory interfaces

Problem

Most of what passes across a screen is gone within the hour. A page read at lunch, a figure in a document, a sentence in a chat: we remember that we saw something, and not where. Search engines index the world. Nothing indexes the part of the world a particular person has actually seen.

Approach

The published application KR 10-2026-0125947 describes a system that provides a search service over a person’s desktop and browser activity. Its classification codes tell the shape of the design: query processing and interfaces, activity logging, natural-language processing, speech recognition, text recognition in images, and generative neural networks. Together they describe a pipeline that records activity, extracts the text and speech inside it, and answers questions about it in natural language.

The research interest for us is less the pipeline than the constraints. Such a system holds the most sensitive record a person can have, so where it computes, what it keeps, and who can ask it questions are the real design problem.

Status

Filed in February 2025 by Suchan An as applicant and sole inventor, published in August 2026, examination pending.