Memory Engineering
Discipline that implements the systems by which an AI agent learns, retrieves, forgets, and evolves information beyond a single session. References the DeepLearning.AI agent-memory course. Sibling of Context Engineering (in-session) and Harness Engineering (infra). Distinct from claude-memory-system (Anthropic-specific concept) and agent-memory-frontmatter (frontmatter mechanism).
Strengths
- Provides a frame to think cross-session learning, beyond frontmatter hacks
- Logical sibling of Context + Harness, coherent discipline of the AI-agents craft
Limitations
- Implementations still very heterogeneous, no cross-runtime standard
Best for
- Agent architects who want vocabulary to discuss retention/forgetting/evolution