Distilled Wiki Pages: Why Origin Composes Memory

A long list of memories eventually becomes another inbox. The agent can search it, but the human has to trust that search will reconcile duplicates, stale facts, and contradictions correctly.

Origin treats storage as the first step. The stronger layer is composition: related captures become wiki pages that explain the current state of a project, decision, workflow, or concept.

Every distilled page keeps the source memory IDs that produced it. Origin's daemon rejects unsourced pages instead of letting attractive summaries enter the store without provenance.

That source chain matters when memory is wrong. You can inspect the original capture, see when it was written, and supersede it instead of guessing why a page says what it says.

A useful memory layer needs to admit that knowledge changes. Origin pages carry revision state and stale reasons so repeated captures can refresh old conclusions instead of silently piling up beside them.

Manual `/distill` is the deliberate path today. Optional local models or API keys can support richer background distill cycles when you want the daemon to do more between sessions.

Agents need compact context, not raw chat archaeology. A distilled page can say the current decision, cite where it came from, and link related entities.

That makes retrieval more useful: Origin can serve atomic memories, pages, graph context, full-text hits, and vector matches together instead of pretending one memory snippet is the whole answer.