ConsilienceKnowledge graph lifecycle (Brain)
Every time you save, it figures out the concepts and relationships in your own sentences and weaves them into a knowledge graph. The agent reasons and works along this ontology. If markdown is your unchanging memory, the knowledge graph is the index that helps you recall what's related. It's a graph that stays alive — learning, linking the same things, clustering related topics — and during downtime the graph touches itself up.
Not a one-and-done — a lifecycle that keeps turning. Learn → Link → Cluster → Tidy.
Every time you save a note, it figures out the concepts and relationships in your own sentences and weaves them into a knowledge graph. It picks up the terms you use in your notes and follows them, and every connection keeps a source pointing back to the note it came from.
It quietly links the same things that showed up under different names. It only lightly connects synonyms, broader and narrower concepts, and related ideas, so the original stays untouched and you can undo it any time.
It automatically groups related topics and draws them as a single topic map. Each cluster gets a name and a summary, and even the bridge concepts that tie several topics together surface — so the big picture of your note collection is visible at a glance.
This is the stage where the graph tidies itself. It learns the terms you use in your notes and feeds them back into the next round of learning, reconciles synonyms, suggests connections worth making, and re-tidies the whole thing or note by note when needed.
The knowledge graph lives in storage on your device and can be exported as files in a portable, standard format. Compare changes at a glance, and rebuild the whole thing from your writing whenever you like.
It doesn't only check whether words match exactly — it follows connections to recall what's related by association. Then it re-gathers what surfaced, so what's truly close rises to the top.
It's tuned so search finishes fast right on your device. It handles even long writing in chunks, and your device's search index covers exact matches too.
It learns the terms and categories you use most and the phrasings you favor in your real notes, then feeds them back into the next round of learning. Terms stop scattering and pull together into one, and older parts get looked after too.
The graph's lifecycle, the topic map, learned terms, synonyms, connection suggestions, and re-tidying. See the whole living process of the graph on one screen, touch it up, and run it again.
Your notes are the originals the system never touches — unchanging memory. The graph, the search index, and the cluster summaries can all be rebuilt. Delete Consilience and your markdown stays exactly as it was.
Just as people learn, index so they can recall, consolidate memory, and tidy up in sleep, the graph goes through the same stages.
Plain markdown files that open anywhere. They're the originals the system never alters, you can move them anywhere, you can compare changes at a glance, and they'll still open in the future.
A knowledge graph layered with an index that helps you recall what's related. It doesn't just check whether words match — it follows connections to recall related things by association.
It quietly links the same things that showed up under different names. The source stays put, and the connection can be undone any time.
Learning your terms, reconciling synonyms, suggesting connections worth making, and re-tidying all keep going. It's why learning isn't a one-and-done thing but a lifecycle that keeps turning.
Local-first. Edit the graph as plain markdown, and the agent reasons on top of it.