Eigenius: A Typed Knowledge-Graph DBMS for AI Scientists

A new open-source database management system, Eigenius, has been introduced to address the limitations of ephemeral scripts in AI-driven research. The system is designed as a typed knowledge-graph DBMS that integrates a dependent type theory, institutions as integration boundaries, and content-addressed immutable storage. Eigenius aims to make data provenance a structural invariant, ensuring that the audit question 'what do you know, and what is your warranty?' can be answered reliably. The system is built on the premise that a unified kernel is necessary for handling stateful, interconnected evidence at scale. The announcement was made via arXiv preprint 2608.04457, with the paper detailing the architecture and its three pillars. The system is positioned as a response to the emergence of 'AI Scientists' that drive research via the Model Context Protocol (MCP). Eigenius enforces epistemic statuses such as declared, observed, derived, and verified, providing a robust framework for knowledge management in AI contexts.