User as Code: Executable Memory for Personalized Agents

2026-06-15 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Expert, quick

Summary

User as Code (UaC) introduces an executable memory paradigm for personalized AI agents, contrasting with traditional unstructured text, knowledge graph, or flat fact stores that rely on retrieval. UaC models a user as a living software project, utilizing typed Python objects for user state and ordinary Python functions to encode governing rules, allowing representation and reasoning within a single, interpretable medium. Its enabling mechanism is a two-phase pipeline: an append-only log that periodically checkpoints into typed code. This approach significantly enhances memory capabilities, matching prior systems on recall (78.8% on LOCOMO) while demonstrating superior performance on aggregate questions over user history, achieving 99% accuracy where retrieval-based methods yield only 6-43%. Furthermore, UaC can surface unsolicited, safety-critical alerts, such as drug-allergy conflicts, a capability absent in query-driven memory systems.

Key takeaway

For AI Engineers designing personalized agents, traditional retrieval-based memory systems often fall short on complex reasoning and proactive alerts. You should explore User as Code's executable memory paradigm to build more robust user models. This approach, using typed Python objects and functions, enables accurate aggregate queries and critical unsolicited safety alerts, significantly enhancing agent intelligence and reliability beyond simple fact recall.

Key insights

User as Code proposes executable memory for AI agents, using Python objects and functions for dynamic, rule-based user modeling.

Principles

• User memory should be executable, not just retrievable.
• Representing and reasoning about user state should occur in one medium.
• Deterministic rule execution enables proactive safety alerts.

Method

A two-phase pipeline: an append-only log captures facts, which are periodically checkpointed into typed Python code representing user state and rules.

In practice

• Aggregate user history queries with high accuracy.
• Proactively detect safety-critical conflicts (e.g., drug interactions).

Topics

• User as Code
• Personalized Agents
• Executable Memory
• AI Memory Systems
• Python Programming
• Knowledge Representation
• Safety-Critical AI

Best for: Research Scientist, AI Architect, AI Product Manager, AI Scientist, Machine Learning Engineer, AI Engineer

Related on AIssential

• Beyond Similarity: Trustworthy Memory Search for Personal AI Agents — Semantic similarity in AI agent memory creates vulnerabilities; a neural gate can ensure trustworthy memory admission.
• Getting Better at Working With You: Compiling User Corrections into Runtime Enforcement for Coding Agents — Compiling user corrections into runtime enforcement rules significantly improves coding agents' compliance with preferences across sessions.
• The man behind Cursor's "memory" feature — Cursor's Memory feature enhances agent context by distinguishing between task-specific and generalizable information for improved coding assistance.
• Agent Memory Patterns in Cognitive Science and AI Systems — Effective AI agents require structured memory systems mirroring human cognition to maintain context and learn.
• MemPro: Agentic Memory Systems as Evolvable Programs — Evolving the entire memory construction-retrieval pipeline as a program, not just the memory bank, significantly enhances long-horizon autonomous agents.
• Agent memory resolved? — A git-like memory framework significantly enhances coding agent performance and context persistence across sessions.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.