A universal Socratic tutor that remembers what you got wrong last week.
LLM
Long-term Memory
EdTech
Socratic Method
//FIG.04 — dissolve
A universal Socratic AI tutor built on a long-term memory architecture that tracks each learner’s progress and misconceptions across sessions. Demoed at the UC San Diego "AI in Higher Education" keynote.
//Role
Creator & engineer
//Status
Beta
//Access
Live / public
//Problem
Most AI tutors forget you the moment a session ends, so they re-teach what you already know and never circle back to what you don’t. Real tutoring is longitudinal.
//Architecture
A Socratic tutoring loop sits on a long-term memory architecture that records each learner’s progress and recurring misconceptions, so the tutor adapts across sessions rather than starting cold every time.
//Role
Creator and engineer — the Socratic engine, the memory architecture, and the learner model.
//Outcome
In beta, demoed at the UC San Diego "AI in Higher Education" keynote in February 2026.
What it took
Technical proof.
▸Socratic tutoring across any subject — questions, not answers.
▸Long-term memory architecture tracks progress and misconceptions per learner.
▸Demoed at the UC San Diego "AI in Higher Education" keynote (Feb 2026).
//Private system — no live link. Architecture and status only; never data, positions, or credentials.←Back to all work
04Beta