A new approach uses personal computer use data to train local LLMs that predict user judgment and writing patterns, reducing the effort required to prompt AI agents. In a two-week study, a specialized model achieved 17.1% semantic accuracy on next-write predictions at $0.3 per call, with a continually trained version reaching 3.0% accuracy at $0.01 per call, suggesting potential for scaling.
An AI engineer built a personalized language learning app to learn Spanish more effectively than existing platforms like Duolingo and Babbel. The app features personalized stories, podcasts, live tutoring, and vocabulary review—all tailored to the user's current knowledge level using comprehensible input principles and AI-generated content.
Tutor-MCP is an open-source adaptive learning engine built on the Model Context Protocol that creates personalized learning paths and AI tutoring for any subject. It uses Bayesian Knowledge Tracing and spaced repetition to guide learners through structured curricula, track progress across AI clients, and maintain persistent learning history.
Spotify introduced a new toggle allowing parents to exclude kids' and family music from their taste profile, Wrapped year-end review, and personalized recommendations. The feature, rolling out globally on mobile for Free and Premium users, addresses long-standing complaints from parents whose listening habits were skewed by children using shared home devices and car speakers linked to parent accounts.
Consumer AI's future will be personalized and context-aware, built on interactive memory systems that users can view and edit together. AI providers will adopt new pricing models based on memory storage alongside subscriptions and token consumption, while innovations like AI avatars, shared canvases, and hardware integrations will expand AI capabilities.
Hightouch launched same-session personalization technology that combines real-time event processing with historical customer data from data warehouses, solving the speed-versus-completeness tradeoff through an incremental compute engine, intelligent caching, and hybrid computation model to enable marketers to act on real-time signals within milliseconds.