mem1.wiki

Patterns

Pattern

Atomic notes

One idea per note, each note independently understandable, linked explicitly. The structural primitive behind Zettelkasten, Evergreen notes, and most modern PKM tools.

Taxonomy
patterns.memory_management
Category
memory_management
Complexity
medium
When to use
Long-running personal knowledge bases where ideas need to compose across years and contexts. Required substrate for graph-shaped retrieval.
When NOT to use
Short-lived project docs, meeting notes, transient capture. Atomising overhead exceeds payoff under ~6 months horizon.

What it is

An atomic note follows three rules:

  1. One idea per note. If you can split it cleanly into two notes, you should.
  2. Self-contained. Reading the note in isolation makes sense — no implicit “as I said earlier” or “see the doc above”.
  3. Explicitly linked. Connection to other ideas is encoded as a link, not implied by physical adjacency in a file.

Andy Matuschak’s “Evergreen notes” is the most cited modern formulation; the original source is Luhmann’s Zettelkasten practice from the 1960s.

Why it’s a pattern, not a method

The unit shape is independent of the philosophy that uses it. PARA can be implemented with non-atomic project docs. Zettelkasten cannot work without atomic notes. The pattern shows up wherever you want notes to compose — to be cited, linked, transcluded, and reused in contexts the author didn’t anticipate.

Where it works

Where it fails

How it relates to retrieval

For RAG and agent memory, atomic notes are the human-curated analogue of good chunking. A vault of well-atomised notes is closer to “pre-chunked” content than a folder of long documents — embedding and retrieval quality both rise sharply.

Sources