This paper explains how AI agents remember things across long conversations and why many current tests don’t truly measure that memory.
Panini is a way for AI to keep learning new facts without changing its brain by storing them as tiny linked Q&A facts in an external memory.