DEV Community

Qonspekt
Qonspekt

Posted on

Using AI to build a Zettelkasten without the friction

The hardest part of building a Zettelkasten isn't the theory. It's consistently turning sources into atomic notes.

Niklas Luhmann wrote 90,000 notes by hand over 40 years. You have articles, papers, and blog posts. The gap between reading them and having them in your vault is where most PKM systems collapse.

What "atomic" actually means

An atomic note has:

  • One concept (not one article, one chapter, or one idea cluster — one concept)
  • A title that could be a Wikipedia article
  • [[wikilinks]] to related concepts
  • A source citation

Bad: "Notes from The Shallows by Carr"

Good: "Neuroplasticity", "Deep Reading Mode", "Hyperlink Distraction Effect" (three separate notes)

The problem with manual atomization

Reading a 4000-word article and extracting 5 atomic notes takes ~25 minutes if you're doing it properly:

  1. Read fully
  2. Identify concepts worth keeping
  3. Title each note (harder than it sounds)
  4. Write the body — in your own words, not copy-paste
  5. Add frontmatter, tags, wikilinks
  6. Add source

Most people skip this entirely or write one large "literature note" that never gets revisited.

How Qonspekt automates this

Qonspekt takes an article and produces atomic notes using Claude AI:

Paste article → Claude extracts concepts → atomic .md files → ZIP download
Enter fullscreen mode Exit fullscreen mode

The system prompt specifically instructs Claude to:

  • Extract 3–7 key concepts (not summaries — named, reusable concepts)
  • Write each as a self-contained note
  • Connect concepts via [[wikilinks]] within the batch
  • Never copy-paste — always paraphrase and synthesize

The wikilink constraint is important. Luhmann's system worked because notes referenced each other. Qonspekt replicates this: the "Spaced Repetition" note from one article will wikilink to the "Active Recall" note from the same article, creating a small interconnected cluster you can drop into your vault.

What to do after importing

  1. Open the ZIP, check the notes
  2. Edit anything that's wrong or imprecise
  3. Add wikilinks to existing notes in your vault ([[existing-concept]])
  4. Move to the right folder

The AI does the scaffolding. You do the integration.

Is this "cheating"?

Depends on your goal. If you're using a Zettelkasten to develop original ideas, the thinking happens when you read and select which articles to atomize, edit the generated notes, and integrate them with your existing vault.

Qonspekt handles the mechanical part. The intellectual work is still yours.

Try it

qonspekt.github.io/qonspekt

BYOK (bring your own Claude API key). No account, no backend. ~$0.003 per article.

Top comments (0)