User as Engram: Internalizing Per-User Memory as Local Parametric Edits

In three linesNovel LLM personalization: store user facts as surgical edits in a hash-keyed memory table (Engram) instead of global LoRA. Reduces memory footprint by 33,000x, improves indirect-reasoning accuracy by 5.6x on average, and enables stacking multiple users without cross-contamination.

Summary generated by Claude — human-verified