Skip to content

Personal Knowledge Management (PKM)

Category: concept Last updated: 2026-04-03 Status: draft

Summary

Personal Knowledge Management (PKM) is the practice of capturing, organizing, connecting, and applying knowledge across a lifetime of reading and learning. The contemporary approach centers on atomic notes, personal knowledge graphs, and increasingly AI skills — turning passive notes into active tools. The core insight: notes trapped in source materials (book margins, isolated devices, disconnected apps) create no leverage; value only appears when ideas are integrated, connected, and made actionable.

Details

The Core Problem

Most people's notes are trapped: - In book margins (can't connect to anything outside that book) - On isolated devices (not synced to the knowledge system) - In apps that don't talk to each other - In a linear structure that obscures the underlying graph of ideas

Notes that stay isolated die in isolation. The value of a note is proportional to how many other notes it can connect to.

The Pipeline (dSebastien's Model)

A 6-step pipeline from analog reading to active AI skills:

Step 1: Capture Anywhere

The system must support capture in any context — at a desk, on a couch, on a train. The medium doesn't matter (paper, reMarkable, iPad, phone). What matters is that everything eventually flows into one system.

Key principle: write in your own words. If you can't explain an idea in your own words, you don't understand it well enough to use later. Copying quotes is not knowledge management.

Marginalia is not knowledge management. Writing in book margins feels productive but creates no leverage — ideas can't connect to anything outside that book.

Step 2: Sync to Vault

All notes must flow into a single, searchable system (the vault). Isolated notes are dead notes.

Tools: reMarkable Sync plugin for Obsidian syncs tablet notebooks directly into the vault as images, preserving folder hierarchy.

Step 3: Convert to Markdown

Handwritten images are converted to searchable, editable text. The Transcriber plugin for Obsidian uses AI to convert handwriting to Markdown and even generates Mermaid diagrams from hand-drawn sketches.

After this step: all notes are digital and structured, but still in a linear block. The real work begins.

Step 4: Decompose Into Atomic Notes

The highest-value step. A book presents ideas in a linear order chosen by the author; the underlying structure of the knowledge is a graph. Decomposing the book into atomic notes recovers that graph.

Atomicity heuristic — connectability: Can you identify links from this idea to other ideas (within the book and across your existing notes)? If yes, it deserves its own note.

Additional questions for each idea: - What can I do with this idea? - What other ideas does this validate? - What does this contradict? - How does this look from a different point of view?

Important: Someone with a richer knowledge graph will extract more atomic notes from the same book — they see more potential connections. This is a skill that develops with practice; a beginner might see 3 standalone ideas where an experienced note-maker sees 10.

Step 5: Connect to the Knowledge Graph

Once atomized, link each note to others — within the same book and across the entire knowledge graph. This is where cross-domain leverage appears: an idea from a habit book connects to something from a neuroscience paper connects to a current project.

Cross-pollination is impossible when notes remain embedded in the linear structure of a single book. The personal knowledge graph is what enables it.

Over time, the graph grows and each new note has more potential connections. The system becomes more valuable the more you use it — compound returns.

Step 6: Turn Actionable Ideas Into AI Skills

Notes are passive. They sit in files waiting to be found. You must remember they exist, find them at the right moment, and figure out how to apply them — three friction points.

AI skills are active. They show up when relevant and apply frameworks to actual problems.

Conversion process: identify actionable ideas (things you can actually DO something with), then ask an AI agent to generate a skill from your notes on that idea. Example from Atomic Habits: "Read my notes about atomic habits. Create a skill that helps me acquire a new habit more efficiently."

Skills compose: a writing skill from one book combines with a persuasion framework from another and a structural thinking model from a third. Each book adds new capabilities to the system.

This is what Agentic Knowledge Management (AKM) looks like in practice.

Atomic Notes

An atomic note captures exactly one idea. Properties: - Can stand alone — understandable without its source context - Can connect — has links to other notes - Can compose — can be combined with other atomic notes to build new understanding

Atomic notes are like LEGO blocks: composable, reorganizable, linkable. The contrast is with linear notes (one big block of text per book chapter) which can't be linked or recombined.

Personal Knowledge Graph (PKG)

The network of connected atomic notes across all domains and time periods. Properties: - Each node is an atomic note - Each link is a meaningful relationship between ideas - The value of the graph grows non-linearly with size (each new note has more potential connections) - The graph represents the underlying structure of knowledge, freed from the linear format of books

The PKG requires a single source of truth — one system where all notes live, so cross-domain connections are possible.

Relation to LLM Knowledge Bases

The concepts/llm-knowledge-base pattern (Karpathy's model) is a related but distinct approach: - LLM knowledge bases focus on external sources (articles, papers, web content) indexed in raw/ - PKM focuses on personal notes synthesized from reading and experience - Both use Markdown and Obsidian as the viewing layer - Both emphasize integration over isolation - The two approaches can compose: personal reading notes become sources in the LLM knowledge base

Key Claims & Data Points

  • Marginalia is not knowledge management — notes trapped in source materials create no leverage — [source: dsebastien_analog]
  • Connectability is the primary heuristic for atomic note boundaries — [source: dsebastien_analog]
  • Someone with a richer knowledge graph extracts more atomic notes from the same book — [source: dsebastien_analog]
  • Notes are passive; AI skills are active — they show up when relevant and apply frameworks — [source: dsebastien_analog]
  • The pipeline compounds: each book makes the system smarter — [source: dsebastien_analog]

Open Questions

  • How do you define "actionable" when deciding which notes to turn into AI skills vs. leaving as passive reference? (raised by: concepts/personal-knowledge-management, 2026-04-03)
  • What happens when AI skills built from different sources conflict — how do you reconcile competing frameworks? (raised by: concepts/personal-knowledge-management, 2026-04-03)
  • At what knowledge graph size does the system start to feel unwieldy — is there an upper limit to useful PKM scale? (raised by: concepts/personal-knowledge-management, 2026-04-03)

Sources