- Semantic search is the #1 pain point — users hate that default search is “barely substring matching” and want meaning-based retrieval like Google
- Smart Connections plugin (4,500+ GitHub stars, 800K+ downloads) shows massive demand for AI-powered semantic search that finds related notes automatically
- Users want proactive surfacing — the tool should show you relevant notes without you having to search
- As you type, show related notes from your past
- “Don’t make me remember tags or search manually — retrieve relevance for me”
- This is the killer feature that new tools like Mem.ai and Constella are building around
- Collaborative note editing and sharing is on Obsidian’s official roadmap
- Teams want to effortlessly discover related project information without manual curation
- Obsidian’s mobile app is praised, but Roam’s mobile “was a flop”
- Quick capture from anywhere, including voice
- Navigation and note retrieval on mobile remains frustrating compared to Apple Notes
- Calendar view for bases/databases is highly requested (on Obsidian roadmap)
- Kanban views for task management
- Better daily notes workflows
- Graph view is popular but often “completely unusable if you have a large number of notes”
- Users want local graphs that update automatically and help navigate connections
- Obsidian’s graph view is praised as “incredible” with filters, groups, and customization
- Roam/Logseq excel here — every block is its own addressable unit
- Obsidian’s block references feel “hacky” with weird identifiers
- People want to link and embed atomic note elements across notes easily
“Every attempt at PKM has landed me in the same place: a huge mess”
- Users collect thousands of notes that never get reviewed or connected
- Quick capture creates massive unprocessed backlogs
- “When there are 200+ notes on a topic, distilling useful information becomes impossible”
- “The more a tool can do, the more it demands from you”
- “You’re not thinking — you’re maintaining”
- Obsidian infamous for the “tinker loop”: install plugins, configure CSS, style graphs, never actually take notes
- “Every feature is another decision”
- Beautiful dashboards that never get looked at again
- “High-maintenance digital gardens that require more watering than they return in fruit”
- Systems that look helpful but waste time
- Constantly switching between Notion, Obsidian, Roam, Bear, Evernote, Anki
- Searching for the “perfect solution” and losing momentum each time
- “Every new note opens up a decision tree: Should this be a page? A subpage? A template?”
- “The last thing you need mid-idea is five technical questions about structure”
- “Most PKMs are great at capture… but almost none are built for recall”
- “I can’t find what I need. Or worse — I forget that I already captured something, because it never surfaces again”
- File system timestamps get changed when copying/syncing
- No official API for accessing Obsidian Sync data
- LogSeq still lacks a sync service
- Natural language queries: “What did I learn about X?” or “Find notes related to this project”
- Mem.ai’s Chat feature lets you “interact with your notes using natural language”
- “Ask questions, edit notes, organize collections, recall details from your knowledge base”
- AI should understand meaning, not just keywords
- Smart Connections plugin uses 1,536-dimensional vectors (OpenAI embeddings) to find semantically similar notes
- Real-time suggestions as you type
- “Heads Up” features that surface related notes while you work
- “Future apps will anticipate information needs based on context, schedule, and work patterns”
- Show relevant past notes without manual search
- 30% reduction in time organizing notes after meetings with AI summarization
- Auto-generate meeting recaps, action items
- Summarize long notes highlighting key points
- AI Note Tagger plugin is trending in 2025
- Auto-categorization and tagging
- Concept mapping and knowledge graph generation
- Voice-to-text transcription for brain dumps
- Capture notes verbally, have them processed into structured text
- AI that requires manual setup/configuration
- Another system to maintain
- AI that doesn’t understand personal context (just general training data)
- Minimal by design: “Only add properties you actively use rather than including unused fields”
- Purposeful metadata: “If a property is not used to answer a relevant question, it should not be included”
- Consistency matters: Use standardized, lowercase property names to avoid duplication
| Property | Purpose |
|---|
aliases | Easier linking (e.g., “Chuck” links to “Chuck Mangione”) |
type | Categorize notes: journal, people, place, organization, recipe |
tags | Organization and DataView queries |
status | Track note state (fleeting, evergreen, discharged) |
created_at | When note was created (filesystem dates are unreliable across syncs) |
project | Project association |
author | For reference notes |
url | Quick access to source |
- Nested YAML: Some users create hierarchical metadata (e.g.,
event.location, event.date) to separate file-level properties from content-level properties
- Type-based templates: Different note types (book, person, event) have different required properties
- Genesis tracking: Track whether note is
original, derived, or external to measure how much is your own thinking
- Adding
creation_date and modification_date when you don’t use them
- Including blank/unused properties in notes
- Over-engineering metadata schemas before you have actual use cases
- Local-first: Plain Markdown files stored on your computer, not in the cloud
- Privacy: No telemetry, no data harvesting — you own your data completely
- Data portability: Notes work in any text editor; you’re never locked in
- Plugin ecosystem: 2,692 plugins, 97.7M+ total downloads in 2025
- Customization: Themes, CSS snippets, configurable everything
- Graph view: Best-in-class visualization of note connections
- Mobile app: Full-featured iOS app with plugin support
- Business model: Free core app, paid add-ons (Sync, Publish) fund development without VC
- Community: 110,000+ Discord members, grassroots growth to 1M+ users
- Pioneered bidirectional linking in mainstream PKM
- Block-level references: Every block is addressable and linkable
- Clean UI: Live preview rendering as you type
- Daily Notes first: Opens to today’s page, captures everything there
- Queries: Powerful embedded queries that surface related blocks
- Task management: Built-in todo cycling with keyboard shortcuts
- Outliner DNA: Perfect for hierarchical, structured thinking
- Best of both worlds: Roam’s outliner UX + Obsidian’s local Markdown files
- Free & open source: No subscription required
- Block references: Same power as Roam
- PDF annotation: Built-in support
- Task states: More granular than Roam (Now/Later/Done)
- Query table view: Clean display for query results
- Privacy: Local storage like Obsidian
| Factor | Why It Matters |
|---|
| Bidirectional linking | Discover relationships organically without upfront organization |
| Plain text foundation | Future-proof, portable, works anywhere |
| Extensibility | Customize to your workflow |
| Active community | Plugins, themes, shared knowledge |
| Philosophy over features | Opinionated about how knowledge works |
- Reduce friction at capture time — zero decisions, zero structure required
- Make recall automatic — don’t make users search; surface relevant content proactively
- AI should understand meaning, not just keywords
- Local-first wins trust — privacy and data ownership matter enormously
- Avoid feature bloat — every feature is a decision users must make
- Block-level granularity matters for serious knowledge work
- The graph view is cool but often unusable — local/filtered graphs are what actually help
- Mobile is still unsolved for most tools
- Don’t make users become janitors of metadata
- The real problem is retrieval, not capture — most tools fail at surfacing past knowledge