Cordelia: Encrypted Pub/Sub for AI Agents

Abstract

We propose a system for persistent, sovereign memory shared between humans and the AI agents they operate. Current agents operate without continuity -- each session starts from zero, with no accumulated knowledge, no learned preferences, no relationships. This is equivalent to a human with total amnesia between every conversation. Cordelia solves this by implementing a distributed memory architecture in which the operator's memory record -- encrypted, replicated, and sovereign -- persists across sessions and gives the agent the continuity needed to function. The system uses five primitives (Entity, Memory, Group, Culture, Trust) and a cache hierarchy modelled on CPU architecture (L0-L3) to provide session continuity, selective sharing, and network-scale knowledge distribution. Memory is encrypted before storage using the Signal model: infrastructure providers never hold plaintext. Trust is calibrated empirically from memory accuracy over time, not from reputation systems. Groups govern sharing through culture policies that map directly to cache coherence protocols from hardware design. The design is deliberately democratic-AI infrastructure: no central provider holds plaintext, no infrastructure operator can coerce the record, and provenance is cryptographically auditable -- properties that matter particularly during the period in which AI capability outpaces the institutional machinery for governing it.