From Evernote to LLM Wiki: Personal Knowledge Management with Zettelkasten and Writing

1. I Had Notes. Was I Using Them?

I have been interested in organizing knowledge for a long time. I started with Evernote, moved to Notion, and later began using Obsidian. After trying different methods in Obsidian, I found Zettelkasten and spent about half a year building notes.

I read and organized whatever caught my interest: economics, history, management, politics, philosophy, psychology, and programming. I ended up with roughly 500 research notes. I was already using AI then. I would ask a question, get an answer or report, and sometimes copy useful excerpts into my notes. At other times, I read and typed the notes myself. There was no single routine.

Over time, the notes accumulated. Search and links made it possible to find related material, but maintaining the system became harder. I sometimes forgot what I had recorded, and rarely returned to notes I had once carefully prepared.

The records were still there. What was less clear was how much they were actually being used in my thinking.

Collecting and connecting knowledge gives us material to think with. To make use of it, we need to choose questions and revisit our records to form judgments. Writing has become one way for me to make that process visible.

2. Looking for a Better Way to Organize

The effort to record knowledge and find it again is older than digital tools. People kept commonplace books of passages and ideas, then developed indexes and categories to retrieve them. In 1945, Vannevar Bush imagined a device that would do more than store documents: it would connect related materials and let people follow those connections in their thinking.

My move from Evernote to Notion and then Obsidian belongs to that same story. I wanted to capture things more easily, organize them more clearly, and find them when needed. The tools changed, but the wish was much the same.

Zettelkasten seemed to offer an appealing answer. If notes were kept small and linked, older records could meet new ideas and help generate further thought. My collection grew, and I enjoyed finding connections across different fields.

But more links did not automatically produce better thinking. A relationship between notes is one thing; finding a question that matters to me in that relationship is another.

I mostly made notes to study. I researched topics that interested me, read, and organized what I learned. I did learn a great deal. Still, I spent less time asking what questions I wanted to answer with that knowledge, or what I wanted to argue or make.

There is nothing wrong with studying for its own sake. Reading and learning can be valuable and enjoyable on their own. In my case, though, organizing knowledge gradually became a goal in itself. I spent time creating new categories and changing the structure of my notes, sometimes without considering what kind of thinking that structure would support. I must have created half a dozen wiki repositories.

My systems changed several times. I am not sure my thinking deepened at the same pace.

A personal knowledge system in which a reader selects useful ideas from sources and connects separate notes around a question

3. What Remains When Management Gets Easier?

Recently, I have been trying a different approach based on Andrej Karpathy’s LLM Wiki proposal. Give an AI materials and a question, and it reads them, creates related notes, connects them to existing records, and updates a knowledge base. It is closer to a working proposal for continuously integrating material with AI than a finished product or a validated theory.

This approach reduces some of the work involved in creating and maintaining notes. AI can help find material, organize it, and connect it. When management becomes easier, another question comes into view:

What will I think about with this knowledge?

The world already has encyclopedias, specialist books, research papers, and public knowledge bases. AI can also explain concepts and show relationships between fields. Its answers still need to be checked for accuracy, sources, and context. Even so, I find myself asking whether reorganizing general knowledge is enough.

Making notes by hand is worthwhile if it helps me understand. Collecting material for a specific problem is worthwhile too. What matters is less the number of records or links than the role they play in my questions and judgments.

Activity	What remains	What to ask next
Collecting and organizing sources	Records and categories	What was I trying to understand?
Connecting notes	Relationships between ideas	What question does this connection raise?
Writing and revising	Arguments and judgments	What changed after I examined the evidence?

Loading diagram.

View diagram source

flowchart TD
  A[Read and collect material] --> B[Choose a question]
  B --> C[Connect records and ideas]
  C --> D[Explain and review in writing]
  D --> E[Refine a judgment]
  E --> B

4. Writing Makes the Shape of an Idea Visible

Since I started writing for JOGYEON, I have moved a step beyond simply saving topics that interest me. A piece might begin with a concept I have been studying, an idea from a video or an old advertisement, or an opinion I want to put into words.

I start by holding on to the subject and reading what I need. I check whether my first explanation holds up and look for missing conditions or opposing views. Then I write.

Vague thoughts become more specific sentences along the way. A premise that seemed obvious in my head may need explanation on the page. After reading more, I may change my initial view or qualify an argument. Writing does more than transfer what I know; it lets me examine the grounds for my judgment.

AI helps organize material, compare explanations, and refine sentences. I choose the questions, review the answers and sources, and decide what belongs in the piece. Compared with my earlier habit of copying excerpts from AI answers into notes, I now more deliberately examine material in light of the question I am writing about.

That does not mean every sentence in a piece began in my head. Ideas change as I work with sources and AI. My questions, review, selection, and revisions are part of that process. A piece of writing puts those judgments where I can look at them again.

Finding a voice does not have to mean inventing an original theory. People notice different passages in the same book. Looking at the same event, one person may focus on its economics while another starts with its history or psychology. A perspective takes shape in the subjects we return to and the evidence we consider important.

Connections between ideas work the same way. I can stop after noticing that two concepts are related, or use that connection to interpret something familiar in a new way. A link records a relationship. Putting that relationship to work in my thinking is a separate step.

Not every piece of knowledge needs to be connected to personal experience. Intuition can reinforce bias, and reviewing evidence may lead me to change an initial belief. The point is not to force personal experience into every topic, but to keep sight of what I am trying to understand.

The notes I made before still inform what interests me. Reading and organizing across fields gave me subjects to write about now. Not every record needs to become an article or a deliverable. Sometimes reading and learning are enough.

I now want to give as much attention to the questions behind my notes as I give to their structure. Writing for JOGYEON is not meant to display knowledge again. It is a way to record what stayed with me from what I read and how I can explain the judgment I reached.

A writing process that uses sources and AI to draft, revisit questions and evidence, and revise the text

From Evernote to Notion, from Obsidian to Zettelkasten and LLM Wiki, I have tried different ways to record and organize knowledge. Lately, I have been writing with that material. Each piece leaves a small record of what catches my attention, what I cannot easily pass by, and which judgments remain after revision.

I do not need to decide too quickly whether to call that my voice. When I write and read the work again, what I keep returning to and what I cannot easily pass over remain between the sentences. For now, I am following those traces into the next piece.