Search papers, labs, and topics across Lattice.
This paper introduces User as Code (UaC), a novel approach to user memory in personalized AI agents that transforms memory from a static retrieval system into an executable model represented as typed Python objects and functions. By employing a two-phase pipeline that logs user interactions and periodically converts them into executable code, UaC significantly enhances the agent's ability to resolve contradictions and perform complex aggregations over user data. The results show that UaC achieves a recall rate of 78.8% on standard benchmarks and maintains near-perfect accuracy (99%) on aggregate queries, outperforming traditional retrieval-based systems, particularly in scenarios where representation and reasoning are critical.
User as Code enables personalized AI agents to execute user memory as live software, achieving near-perfect accuracy on complex queries while surfacing critical safety alerts.
A personalized AI agent needs a user memory: a persistent model of who the user is, built across many conversations and consulted on each new one. Today this memory is almost always stored as unstructured text, a knowledge graph, or a flat store of facts, and consulted by retrieval -- fetching the entries most similar to the current request. Such "bag-of-facts" memory recalls individual facts well, but because storing a fact and acting on it are separate steps, it struggles to resolve contradictions, aggregate over many records, or enforce rules. We argue that user memory should instead be executable. We introduce User as Code (UaC), a paradigm in which an agent's model of a user is a living software project: typed Python objects hold the user's state and ordinary Python functions encode the rules that govern it, so representing and reasoning about the user happen in one medium an interpreter can run. The enabling mechanism is a two-phase pipeline: an append-only log that never discards a fact, periodically checkpointed into typed code. This changes what memory can do. On standard long-term conversation benchmarks, UaC matches both a full-context upper bound and the strongest prior memory systems on recall (78.8% on LOCOMO). Its advantage emerges where representation matters most. On aggregate questions over a user's history -- "how many international trips did I take last year?" -- retrieval-based memory collapses (6-43%) while UaC stays near-perfect (99%), because the answer is a one-line computation over typed state rather than a search over text. And because its rules execute deterministically whenever the state changes, UaC can surface unsolicited, safety-critical alerts -- such as a newly prescribed drug that conflicts with an allergy recorded months earlier -- a capability query-driven memory cannot provide.
