Alex' Gardenアレックスの庭

5 min read

Standalone PKM app — cons of Obsidian setup, sync, UI focus (2026-05-06, 14:03)

First memo in a same-day trio extending OpenClaw for PKM toward a standalone app. Sister memos: 2026-05-06-Standalone PKM app - Cursor SDK as harness layer (14:21) and 2026-05-06-Standalone PKM app - yjs CRDT sync stack (15:06). Frames the gap (current Obsidian + Claude Code rough edges) and the design pull (purpose-built UI for ingestion + retrieval, not manual editing).

i want something like my current obsidian AI setup, but bundled as a separate app so that it doesnt require all of the rough edges obsidian has; current cons of obsidian / claude setup:

  1. steep learning curve → only enthusiasts use it
  2. sync has to be managed by the user
  3. relies on git, which is cumbersome and not built for many small changes (especially on mobile it doesnt really work too well)
  4. setup is clustered with many moving parts and the user needs to integrate it themselves (e.g. git repo, plugins, claude code, claude code routines …)

other potential improvements a standalone app would bring:

  • better ai performance (e.g. using vector db embeddings)
  • more focused experience
  • purpose built UI: not focus on manually editing notes (like obsidian has this focus) but rather focus on ingestion (e.g. provide quick notes feature, voice memos etc) + retrieval via AI + notification center or similar

challenges:

  1. ai provider cost - user would have to provide their own AI provider (maybe this can be fixed with provide api key or login with codex/claude?!); best option would be to offer plans but i currently dont want to build paid apps due to tax complications
  2. time/resources: building such an app will probably take more than just building a little calendar like hibi
  3. local first md vs databases?! most integrated ai apps like e.g. fabric.so aren’t really local first; however being local first is probably a good idea since it has so many other benefits + large community?! how to balance this?
  4. UI: what would this new UI focus look like? how to balance normal text editor (like traditional PKMs) with the “Chat only” UI? Also: chat uis are so popular right now, just showing a chat UI would probably turn most users away since it makes the app look like just another AI app…
  5. refine / optimize the core loop: is the current setup enough / scalable?
  6. cloud routines: would essentially have to re-implement claude code routines with own infra - sounds like a huge challenge to get right TBH

ideas:

  • use git as sync engine with backend server having an integrated git engine or well using some sort of sync service that is essentially just a lightweight git server with auth per user?! security?!
  • or well just leave out git completely and have an own local file based sync engine in place (maybe just keep it apple native for now and use containers or similar?! - however how to do cloud routines then - seems impossible with that apple native sync layer)

Graph View

Backlinks