Random Walk

Luhmann deliberately created “accidents” by revisiting old cards. This skill operationalizes that habit.

Inputs

• mode (default: random): random / theme / time / cross
• count (default: 4): recommended 3–5
• filter (optional): keyword or time prefix (e.g. 202401)

Workflow

Step 1: Sample notes

Sample from your note directories (adjust to your repo):

• Notes/
• Literature/

Strategies:

• random: slightly bias toward older notes
• theme: filter by keyword, then random
• time: filter by date prefix
• cross: force one note from different domains/categories

Step 2: Deep read — “three link questions”

For each sampled note:

1. How does it relate to my current problem/project?
2. What does it echo from recent learning?
3. What new question does it create?

Step 3: Action output

• New link → propose bidirectional write-back edits
• New idea → propose a new fleeting/permanent note
• Needs update → specify what to refresh
• No signal → record why and adjust next parameters

Output template

## Random Walk Log [YYYY-MM-DD]

### Sampled notes
1. [[Note 1]] — one-line takeaway
2. [[Note 2]] — one-line takeaway

### Serendipity insights
- **[Note A] × [Current project]**: ...
- **[Note B] × [Note C]**: ...

### Actions
- [ ] [Concrete write-back change / new note]

Quality checks

• ≥2 bidirectional link suggestions that can be written back
• ≥1 cross-domain analogy or “unexpected link”
• If no discoveries: explain why + propose better parameters