Should agent memory model facts, episodes, and procedures as separate types?
Fun-Following-1723 · reddit · 2026-08-28
A junior developer building agent memory systems asks a practical design question: beyond scoping memory into user / agent / shared-project, should memory type be its own dimension—fact ("user prefers English"), episode ("user requested leave"), and procedure ("HR agent follows workflow X")?
The intuition is that types behave differently at retrieval: a preference is useful across many agents, while an old episode shouldn't be injected unless relevant to the current task. The question to practitioners: do you model type separately from scope, or keep one memory kind with metadata and let retrieval handle the difference—and does memory type actually change your retrieval/injection strategy in production?
More from coding & agent
- AQuA paper: if an agent can edit its tool schema, you're not evaluating the same system — creditme7 · 2026-08-28
- 4 Hours on a Train: A Fold Plus Remote Desktop Is Enough to Run All Your Agents — kevinkern · 2026-08-28
- Accio Open-Sources CommerceAgentBench for Real Agent Tasks — kimmonismus · 2026-08-28
- Code World Model: A Coding Agent as World Brain, Code for Rules, Video for Rendering — burny_tech · 2026-08-28
- MIT-licensed tools reduce coding agent context waste without secondary LLMs — bonsaisushi · 2026-08-28
- Testing IRL evaluation method saves 20-70% compute cost — xeophon · 2026-08-28