Palinode builds on ideas from across the agent memory landscape:
- Karpathy's LLM Knowledge Bases — the original "agents need memory" gist that kicked off much of this work
- Letta — tiered memory (main context + archival), self-editing memory blocks
- LangMem — typed schemas, background consolidation
- memsearch — hybrid BM25 + vector search, content-hash dedup
- Hermes — FTS5 sanitization patterns, security scanning
- OB1 — two-door capture pattern (structured vs unstructured intake)
If you know of prior art we missed, please open an issue.