Skip to content

drewburchfield/kindling

Repository files navigation

kindling

License: MIT

A writing and knowledge management system that blends Niklas Luhmann's Zettelkasten, Tiago Forte's CODE (Capture, Organize, Distill, Express) and PARA (Projects, Areas, Resources, Archive) with Claude Code as the discipline layer.

The foundation is Zettelkasten's atomic philosophy: capture a three-word phrase, a single sentence, or a thousand-word rant. The system treats all of it as valid raw material. No structure required upfront, just get it out of your head. That flexibility is the core value.

Layer on Claude Code + CODE methodology and things actually get processed (most people's inbox is where ideas go to die). Layer on PARA and processed knowledge feeds into real projects and areas of your life.

Built on Obsidian + semantic graph search.

What You Can Do

  • Process your inbox into atomic notes with AI-assisted triage and insight extraction
  • Discover hidden connections between notes across completely different domains
  • Search your knowledge with semantic search that adapts depth to your query
  • Think through problems with a Socratic partner that resists jumping to solutions
  • Preserve your voice with AI staging, human rewriting, and de-AI-ification tools
  • Track your intellectual evolution with daily reviews, weekly synthesis, and changelogs

Quick Start

See INSTALL.md for full setup. The short version:

  1. Clone this repo and open it as an Obsidian vault
  2. Configure MCP servers (obsidian REST API + obsidian-graph)
  3. Copy config templates and update vault paths
  4. Start Claude Code from this directory
claude

Test it:

/recall test

Getting Started

Week Focus Commands
1 Capture. Dump thoughts, articles, notes into 00_Inbox/ /process-inbox
2 Connect. Run connection analysis on permanent notes /find-connections, /update-changelog
3 Discover. Look for cross-domain patterns /discover, /recall
4 Reflect. Establish your rhythm /daily-review, /weekly-synthesis
When What Time
Throughout the day Capture to 00_Inbox/ Seconds
Evening /daily-review 5 min
Weekly /process-inbox + /weekly-synthesis 30 min
Monthly /analyze-kb + /discover 1 hour

See EXAMPLES.md for full workflow walkthroughs.

Commands

Your state of mind Command What it does
"I have stuff piling up" /process-inbox Triage items one at a time, extract insights to staging
"I just added a note, what connects?" /find-connections <note> Map the conceptual network around a specific note
"What patterns am I missing?" /discover Vault-wide cross-domain pattern hunting
"What do I already know about X?" /recall <topic> Semantic search, adapts depth to query complexity
"Help me think through X" /thinking-partner <topic> Socratic exploration, resists jumping to solutions
"Make this sound like me" /de-ai-ify <file> Remove AI voice patterns, restore human language
"What did I do today?" /daily-review End-of-day capture and reflection
"What happened this week?" /weekly-synthesis Weekly patterns, energy audit, theme detection
"Log what was discovered" /update-changelog Record connections and decisions from this session
"Switch vaults" /switch-brain <path> Update all config files for a different vault

Plus /analyze-kb for vault health reports.

Workflow Chains

Processing new content:

/process-inbox -> human review -> /find-connections -> /update-changelog

Periodic discovery:

/discover -> /update-changelog

Preparing to write:

/recall <topic> -> /thinking-partner <angle> -> write -> /de-ai-ify

How It Works

Three methods, layered:

Layer Method Role
Foundation Zettelkasten The engine. Atomic insights that connect and compound over time.
Workflow CODE The nudge layer. Tells you what phase you're in: Capture, Organize, Distill, Express.
Structure PARA The collection layer. Predetermined areas where processed work lands: Projects, Areas, Archive.

The core question for every piece of content: "Is this building knowledge or managing action?"

  • Building knowledge: Sources -> Permanent Notes -> MOCs -> Output
  • Managing action: Projects (deadline) or Life Areas (ongoing)

AI Assists, Human Writes

AI-extracted content never goes directly into your permanent notes. It goes to AI_Extracted_Notes/ tagged with #ai-extracted. You review it, rewrite it in your voice, then promote to 02_Permanent/.

What Makes /discover Different

Most knowledge tools find similar content. /discover does the opposite: it hunts for notes with low semantic similarity but high conceptual connection. A note about dopamine dynamics and one about prospect theory might score 0.63 similarity, but they describe the same mechanism (baseline-deviation adaptation). That's the kind of connection that sparks an article.

Folder Structure

00_Inbox/           # Dump everything here (flat, no subfolders)
01_Sources/         # Books, articles, writing exercises
02_Permanent/       # Atomic insights, one idea per note (THE CORE)
03_MOCs/            # Maps of Content (navigation hubs)
04_Projects/        # Time-bound work with deadlines
05_Life_Areas/      # Ongoing responsibilities
06_Output/          # Synthesized articles and frameworks
07_Archive/         # Completed work
08_Attachments/     # Media files
09_Meta/            # System files, changelogs, templates
AI_Extracted_Notes/ # AI-generated content (clear provenance)

Only the framework is tracked in git. Your actual notes sync however you prefer (iCloud, Dropbox, etc.). Each folder has a README explaining what belongs there.

Permanent Notes

Every permanent note is one atomic insight, written in your words. 50-300 words, clear title that states the insight, [[wikilinks]] to related notes, YYYYMMDD prefix.

---
created: 2024-12-16
tags: [neuroscience, motivation]
---

# Dopamine drives wanting, not liking

Dopamine is responsible for the *desire* to act, not the enjoyment
of the action itself. This explains why we can be motivated to do
things we don't enjoy.

The "wanting" (dopamine) system is separate from the "liking" (opioid) system.

**Source**: [[Dopamine Nation - Book Notes]]

**Related**:
- [[Motivation vs willpower]]
- [[Reward prediction error]]

Requirements

MCP Server Purpose Required
obsidian Read/write files in your vault via Obsidian's REST API Yes
obsidian-graph Semantic search, connection graphs, hub/orphan detection (pgvector) Yes
Docker MCP gateway YouTube transcript pulling and other utilities Optional

License

MIT License - see LICENSE for details.

About

Turn raw thoughts into written work. Blends Zettelkasten, CODE, and PARA with Claude Code as the discipline layer. AI thinks with you, extracts and organizes. You write.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors