Students lose continuity with AI tutors — persistent user profiles & memory

Learners, instructors, and institutions increasingly rely on LLM-driven tutoring, but sessions are typically stateless and students lose continuity: gaps in prior explanations, forgetting past misconceptions, and fractured progress tracking that undermine retention and long-term outcomes. This affects millions of users across K–12, higher ed, and lifelong learners; at a macro level there are roughly 1.25 billion learners globally, representing an addressable digital-personalization attach market of about $12.5B (at $10/yr per learner). A viable product is a persistent-user-profile and memory layer that augments real-time LLM tutors with a longitudinal skill graph, selective episodic memories, and RAG-enabled personalization stored in hosted vector DBs. Sell to institutions via outcomes-tied contracts and to consumer apps via per-user subscriptions; the concept maps to the listed revenue potential (90/100) but requires solving privacy, consent, cold-start, and integration challenges. Timing is favorable: LLMs are now accurate and cheap enough for real-time tutoring and vector stores are commoditized, while buyers increasingly demand outcome-based procurement, which supports a market score of 95/100. To stand out in a medium-competition field, focus on measurable longitudinal impact (validated retention lifts), tight LMS/assessment integrations, compliance (FERPA/GDPR), and defensibility through proprietary longitudinal learning signals and outcome-based pricing; the main risks are data governance complexity and the need to demonstrate causality, not just correlation.