Chain of Thought | AI Agents, Infrastructure & Engineering Agent Memory: The Last Battleground in the AI Stack | Richmond Alake
13 snips
Apr 2, 2026 Richmond Alake, Oracle's AI DevEx lead and creator of the MemoRiz library, is a leading voice on agent memory. He describes why memory engineering deserves its own discipline. He demos a memory-aware financial agent that runs vector, graph, spatial, and relational search in one query. He explains controlled forgetting, four human memory types mapped to agents, and why databases beat files in production.
AI Snips
Chapters
Transcript
Episode notes
Memory Engineering Is Its Own Discipline
- Memory engineering is a distinct discipline combining database, data, software and agent engineering focused on systems that learn, remember, and adapt.
- Richmond frames memory engineers as those who design retrieval pipelines, data models, and memory-aware systems rather than just writing prompts.
Design Agents With Memory First
- Think memory first when designing agents and treat memory as a first-class primitive rather than an afterthought.
- Add attributes to memory units that allow recall and controlled forgetting to support long interactions and audit requirements in production.
Don't Delete Data Use Controlled Forgetting
- Avoid deleting stored information; implement forgetting via decay so you can audit and recover historical data.
- Richmond's rule: 'Don't delete, forget' — keep records and reduce relevance scores rather than removing entries outright.