The Forgotten Art of Knowing What You Know

Niklas Luhmann published 50 books and over 600 articles in a career spanning four decades. When asked how he was so productive, he didn't mention discipline or an exceptional memory. He said he worked in partnership with his Zettelkasten.

Not "I used it as a tool." A partnership. That word choice is doing something.

The Zettelkasten — German for "slip box" — was a wooden cabinet containing roughly 90,000 handwritten index cards, each with a single idea, each linked to others by a numbering system Luhmann invented himself. He started it in the 1950s and kept at it for forty years. By the end, he described it as a conversation partner. He'd write a question into the system, follow the links, and ideas from years earlier would surface in new configurations. He credited it not with storing his knowledge but with generating it.

That's worth sitting with. The system wasn't a filing cabinet. It was a structure for thinking.

The discipline that's been forgotten

Before productivity software, before the cloud, maintaining personal knowledge was treated as an art form with formal instruction. Commonplace books — notebooks where thinkers collected quotes, observations, arguments, and examples — were standard practice among educated Europeans from the Renaissance onward. John Locke designed a formal indexing scheme for his commonplace book and included it as an appendix to his Essay Concerning Human Understanding, as if knowing how to organize one's reading was as important as the reading itself.

Leonardo da Vinci kept notebooks throughout his life. In one, he described them as "a collection without order, drawn from many papers, which I have copied here, hoping to arrange them later." Even Leonardo planned to do it properly eventually. He never did, by most accounts.

The Zettelkasten formalized what the commonplace book had done intuitively: build a system that could surprise you. The goal wasn't retrieval in the sense of finding what you'd stored. It was retrieval in the sense of finding what you'd forgotten you knew, and watching it connect to something new.

David Allen's Getting Things Done, published in 2001, restated the core principle for the digital era. His central claim was blunt: "there is an inverse relationship between things on your mind and those things getting done." The point of a trusted external system wasn't efficiency. It was cognitive freedom. You capture things precisely so you don't have to hold them.

What modern tools got wrong

The explosion of note-taking software over the last fifteen years produced something that looks like those systems but works differently. Notion, Roam, Obsidian, Bear — each made capture faster and retrieval harder.

Not harder technically. Harder cognitively. The paradox of digital PKM is this: the system was built to offload memory, but navigating it efficiently requires memory. You need to remember that you stored something, remember roughly when, remember what words you used at the time. The system ends up asking you to maintain a parallel mental index of itself, which defeats the purpose.

The commonplace book didn't have this problem because it was small enough to browse. The Zettelkasten solved it with links — you didn't need to remember where something lived, you navigated there from somewhere adjacent. Modern tools added full-text search and called it solved. It isn't. Full-text search finds words. It doesn't find thoughts.

There's also a discipline problem. The old systems required regular review, active linking, and deliberate curation. That friction was the point. Luhmann didn't dump ideas into his slip box. He wrote each one as a complete thought, placed it carefully, and linked it manually to everything it touched. The discipline of doing that was the discipline of thinking.

Most people who set up an Obsidian vault in 2023 skipped that part.

Whether AI changes the calculus

Here's where I'm genuinely uncertain.

The optimistic case: AI can do the linking work that made the Zettelkasten powerful. It can surface connections across thousands of notes you'd never find manually, update relationships as context changes, retrieve not just words but the concept behind a query. That would be a real improvement over both the old systems and the digital tools that failed them.

The pessimistic case: AI might just make capture faster and retrieval smoother without touching the underlying problem, which is curation. Knowing what to keep. Knowing what connects. Knowing what to discard. Luhmann spent forty years curating 90,000 cards. The system worked because he thought carefully about each one. If AI handles capture automatically and retrieves on demand, what does the human actually learn to know?

Probably the right answer is somewhere uncomfortable: AI is genuinely useful for retrieval and surface-level connection, but the parts of the old systems worth preserving are precisely the parts that require effort. The review. The deliberate linking. The decision about whether something belongs or should be thrown away.

Luhmann called it a partnership because both sides contributed. The system did the remembering. He did the thinking. Separating those two jobs is what made both possible.

Most of us are trying to outsource both.

---

Asgeir Albretsen is the founder of Harbor.