Second Brain

Lesson 5 of 5

Discovering Connections

Estimated time: 5 minutes

Discovering Connections

Here's where your Second Brain becomes more than just search. You've been saving notes from different contexts — conferences, books, work meetings, articles. But ideas don't respect categories. The hiring advice from a management book might connect to the team dynamics problem in your project retro. In this lesson, you'll configure OpenClaw to surface these hidden connections automatically.

Prerequisites

    Connection Architecture

      Knowledge Base                 Connection Engine
      ┌────────────────────┐        ┌──────────────────────────┐
      │ Conference notes   │        │                          │
      │ Book highlights    │───────>│  Cluster similar chunks  │
      │ Work decisions     │        │  Map cross-source links  │
      │ Article clips      │        │  Detect theme patterns   │
      │ Podcast quotes     │        │  Generate insights       │
      └────────────────────┘        └──────────┬───────────────┘
                                               │
                                        Weekly "connections"
                                        digest in your chat
    

    Enable Connection Discovery

    Turn on the connection engine, which runs periodically in the background to find relationships between your notes.

    connections-config.yaml
    connections:
    enabled: true
    schedule: weekly           # Run every Sunday evening
    min_similarity: 0.78       # Threshold for "related" content
    cross_source_only: false   # Also find connections within same source
    max_connections_per_run: 20digest:
    enabled: true
    channel: direct_message  # Send to your DM, not group chat
    day: sunday
    time: "18:00"
    max_items: 5             # Top 5 most interesting connections
    CLI
    openclaw knowledge connections enable --schedule weekly

    Read Your First Connections Digest

    After the engine runs, you'll receive a digest like this.

    Weekly connections digest
    Bot:  Your Second Brain Connections Digest — Week of Mar 3
    
    UNEXPECTED LINK
    Your note from "Shape Up" (book) about "appetite-based
    planning" connects to your retro note from December
    where the team struggled with "scope creep on the API
    project." Shape Up's fixed-time/variable-scope approach
    directly addresses the problem your team identified.
    → View both notes
    
    
    RECURRING THEME
    The phrase "developer experience" appears in 8 notes
    across 4 sources (2 conference talks, 1 book, 1 article).
    You seem to be building expertise here. Want me to
    create a synthesis note?
    → See all 8 notes
    
    
    CONTRADICTING VIEWS
    Your note from the CTO's talk says "monorepos are the
    future" but your saved article from the Uber eng blog
    argues "monorepos don't scale past 500 engineers."
    Interesting tension worth exploring.
    → Compare both perspectives
    
    
    KNOWLEDGE GAP
    You have 6 notes about "event-driven architecture" but
    they all cover theory. No notes about implementation
    or real-world experience. Consider saving some practical
    guides or case studies.
    
    
    FADING KNOWLEDGE
    Your detailed notes from the Leadership Summit (9 months
    ago) haven't been referenced since. Key insight that
    might be worth revisiting: "Psychological safety is the
    #1 predictor of team performance — more than skill."
    → Review these notes
    
    Reply with a number to explore any connection.

    Act on Connections

    The digest isn't just interesting — it's actionable. When the bot flags a connection between a book concept and a work problem, that's your cue to apply the insight. Reply "Tell me more about #1" to get a deeper analysis.

    On-Demand Connection Queries

    Don't wait for the weekly digest. Ask for connections anytime.

    Connection queries
    You:  What in my notes relates to the hiring challenges
    we discussed last Tuesday?Bot:  Your Tuesday meeting note mentions three challenges:
    "Can't assess culture fit in 45-min interviews"
    "Senior candidates want remote, we're hybrid"
    "Technical interviews don't predict job performance"
    Related content from your other notes:For challenge #1:
    "Who" book notes: "Use structured scorecards with
    behavioral questions tied to specific outcomes"
    Podcast note (Lenny's Podcast #89): "Trial projects
    beat interviews for culture assessment"
    For challenge #3:
    Article clip (from interviewing.io): "Take-home
    assignments predict job performance 3x better than
    whiteboard coding" — with data from 5,000 interviews
    Want me to compile these into a brief for the team?

    Build Knowledge Maps

    Visualize how your knowledge clusters together.

    Knowledge map
    You:  Show me a map of my knowledge clustersBot:  Your knowledge base has 6 major clusters:ENGINEERING (342 chunks)
    ├── Architecture patterns (89)
    ├── DevOps & infrastructure (67)
    ├── Programming languages (54)
    ├── Testing & quality (42)
    └── Performance & scaling (38)LEADERSHIP (198 chunks)
    ├── Team management (72)
    ├── Hiring & recruiting (45)
    ├── Communication (38)
    └── Strategy & planning (43)AI & ML (156 chunks)
    ├── LLMs & transformers (67)
    ├── RAG & embeddings (34)
    └── AI safety & ethics (55)BUSINESS (94 chunks)
    ├── Product strategy (42)
    └── Fintech & payments (52)PERSONAL DEVELOPMENT (67 chunks)
    HEALTH & FITNESS (31 chunks)Cross-cluster connections: 47 links
    Strongest bridge: Leadership ↔ Engineering (23 links)
    Weakest area: Health & Fitness (few connections to other topics)

    Cluster Visualization

    Run openclaw knowledge map --export html to generate an interactive visual map you can explore in your browser. Nodes are chunks, edges are connections, and clusters are color-coded.

    Want random inspiration? Enable serendipity mode for occasional nudges.

    Serendipity config
    serendipity:
    enabled: true
    frequency: daily
    time: "09:00"
    style: "brief"            # One quick note, not a full digest

    Each morning, the bot surfaces one random note you might have forgotten: "From 8 months ago — you highlighted this in 'Thinking in Systems': 'A system's behavior is determined by its structure, not by external forces.' Still relevant?"

    It's like a daily flashcard from your own knowledge.

    Some connections are too valuable to keep to yourself. You can share specific connections to a team channel:

    Sharing
    You:  Share the connection between the Shape Up notes
    and our API retro to the #engineering channelBot:  Shared to #engineering:"Interesting connection from my notes — the scope
    creep issue from our API retro maps directly to the
    'appetite-based planning' concept in Shape Up.
    Key idea: fix the timeline, flex the scope.
    [Link to Shape Up summary] [Link to retro notes]"

    This only shares the specific connection and quotes — not your entire knowledge base.

    Knowledge Connections API

    API reference for programmatic access to connection data, cluster maps, and digest customization.

    https://docs.openclaw.ai/knowledge/connections

    Knowledge Check

    What type of connection is most valuable in a Second Brain system?