Persistent Synthesis Memory

Persisting cross-source synthesis lets agent knowledge compound instead of reconstructing the same relationships at every retrieval.

Conventional retrieval brings raw fragments into each query and asks the model to reconstruct their relationships anew. A persistent synthesis memory instead stores agent-authored concepts, cross-references, contradictions, and revised conclusions between sessions. New sources update the existing synthesis, so earlier integration work becomes durable input to later questions.

This is more than indexing and more mutable than the raw evidence. Preserve the distinction with source-memory-schema separation, carry provenance into derived notes, and run an explicit ingest-query-maintain loop so accumulated synthesis does not become accumulated staleness.

---

Sources

• LLM Wiki, The core idea