Self-Learning Agent — Knowledge Card Memory System

A production-tested memory architecture for AI agents that wake up fresh each session. Instead of one monolithic memory file that grows until it's unusable, this system uses atomic knowledge cards (~350 tokens each) searched semantically, daily logs for raw notes, and a slim master index loaded every session.

Architecture

workspace/
├── MEMORY.md              # Master index (~2KB, loaded every session)
├── memory/
│   ├── cards/             # Knowledge cards (~350 tokens each)
│   │   ├── topic-name.md  # One topic per file, YAML frontmatter
│   │   ├── another-topic.md
│   │   └── ...
│   └── YYYY-MM-DD.md      # Daily session logs (raw notes)

Why This Works