Sato

Knowledge should belong to the learner.

Sato brings together personalized learning, local knowledge retrieval, and a student-owned agent runtime. The starting point is a question about ownership: where should the memory of your learning live?

A notebook is a record of how someone thinks. A learning system should be able to work alongside that record without taking it away from them.

The current architecture

The conceptual architecture keeps textbook knowledge local: embedded chunks, represented as 768-dimensional vectors in ChromaDB, sit alongside a local agent runtime. Notebook integration connects retrieval to the place where the student is already working.

Local inference and decentralized infrastructure are part of this direction. The work is in connecting these pieces into a learning layer that remains understandable to the person using it.

Still working through

How can a system help a student make a connection without making the connection on their behalf? Retrieval is only part of the problem. The more interesting part is what happens after something has been found.