AI Native PKMs

==tl;dr: AI Native PKMs, using a Graph DB to store knowledge==

Why?

• old learning methods are outdated
• i find myself mostly using chatgpt in combination with Obsidian / Anki
• i wish there was a way to do all the learning in chatgpt
• goal: have a chat that you can use to learn new stuff and automatically update the saved knowledge (like automatically creating obsidian notes etc)

What?

• views:
  • chat view (main)
  • knowledge base
    • stored in a graph db
    • graph UI + tabular UI

Open Questions

• what should the graph ontology look like?
• how to ensure fast responses in the chat view, even tho a lot of RAG (and updating) is necessary?

How?

• webapp
• neo4j?!

Related

• OpenClaw for PKM — proactive AI background agent; the backend layer this envisions
• Open-Source Model-Agnostic AI Platform — the full platform framing: open-source Claude equivalent (chat + coding agent + routines)
• Agent-Native Software Architecture — the architectural pattern underlying this kind of AI-native tooling
• 2026-05-06-Standalone PKM app - cons of Obsidian setup, sync, UI focus — recent concrete attempt at this idea, listing the specific pain points an “AI-native PKM” would need to solve.
• Kumo — storage substrate this kind of PKM could sit on; agent-first cloud filesystem with MCP/CLI surface, no opinionated PKM structure (“pure storage + tooling”) so a PKM app brings its own workflow.
• 2026-05-19-Distill app review - AI-native PKM with local-only limits — shipped instance of this idea: chat + atomic chat messages (replacing the older graph-DB ontology question), vector-indexed, end-to-end encrypted sync. The 5/19 review surfaces what works (the atomic-unit choice, lists instead of folders, BYO-model endpoint) and what doesn’t (local-only agents/MCP, trust, no extensibility).