We propose the Personal Small Model (PSM), a novel architecture for AI agent memory inwhich a small, per-deployment model is trained not to store user content, but to master memoryoperations: consolidation, decay scheduling, recall weighting, interference detection, and sleep-timereorganization. Unlike existing approaches that treat memory as a retrieval problem—injectingdatabase fragments into a language model’s context—the PSM treats memory as a learned cognitiveskill, architecturally separated from the primary reasoning system. The PSM’s weights remainshared and stable across all users; personalization lives entirely in per-user memory stores that thePSM manages. This design eliminates catastrophic forgetting by construction, enables biologicallyinspiredmemory consolidation, and allows a large language model to benefit from rich personalcontext without any modification to its architecture. We present the full system design, trainingmethodology, memory tier hierarchy, and a sleep-time consolidation algorithm. This documentconstitutes a public prior art disclosure. No patent is sought.