The Zettelkasten Trap
Why Most PKM System Collapse Under Their Own Weight
800 notes and zero output. Sound familiar?
That was my Zettelkasten trap.
The problem with Zettelkasten and other PKM methods is just how easy they make capture feel. Read something interesting, clip the highlight, write a quick note, add a tag, maybe link it to another note. Each action makes the system feel stronger.
But after a while, I had to ask the uncomfortable question one writer put plainly: “I have 4,000 notes in my vault. Cool. What came out of them?”
I started seeing versions of that question everywhere: impressive vaults, elaborate graphs, careful tags, and very little to show for it. So it wasn’t just me.
A note system only becomes useful when it reliably turns captured ideas into retrieval, reflection, and output.
The common assumption is that if we capture enough, inspiration will eventually arrive. More notes means more raw material, more connections, and eventually more output.
That story makes sense. It also lets capture become the work.
Capture is the easiest part of knowledge work. It asks very little from us. We notice something, preserve it, and move on. Retrieval, reflection, and output are harder because they require decisions. What does this mean? What am I going to make from it?
The internet already beats most personal note systems as a reference library. AI has made that even more obvious. What they don’t have is my trail of attention: the ideas I found surprising, useful, frustrating, or personal over years of reading and working.
The reframe I’m working with now is simple: a note system is a pipeline, not a storage bin.
If the note is the final resting place of an idea, the system is mostly an archive. Archives are useful, but they are not working systems. A working system has movement. Something enters, is processed, revisited, and eventually becomes a decision, outline, draft, project, or conversation.
David Allen’s GTD describes a full loop: capture, clarify, organize, reflect, engage. Most PKM conversations spend a lot of time on capture and organization. Those are important, but they are not the whole loop.
For my own system, I think about the back half of the pipeline in three stages: retrieval, reflection, and output.
Retrieval asks whether I can find the right note at the right moment. Reflection is where raw notes become thinking again. Output is the honest metric: did the system help me write, decide, or plan?
My Zettelkasten had almost none of this.
I captured constantly and reviewed rarely. I had a lot of notes, but few projects pulling those notes forward. Retrieval was mostly theoretical. I had built a place for knowledge to accumulate without building enough pressure for that knowledge to become anything.
So the current iteration has not been “find a better capture method.” It has been adding the missing back half of the pipeline.
The first step was consolidation. Over the years, I had notes scattered across different systems and formats. Bringing them into one place made retrieval immediately more realistic. Once the notes lived together, I could use Emacs, consult, and ripgrep to search the whole collection. That alone changed the system for me.
The second step was reflection, and that is where the experiment got more interesting. One idea I took from Sonke Ahrens’ How to Take Smart Notes is that notes should become more atomic over time. My problem was that I had no reliable way to see which notes needed that treatment.
This is where I began using AI in the review process.
With help from Claude Code, I worked on getting my notes into a structure that AI could analyze by concept, not just by tag. I processed them using Voyage AI embeddings and added them to ChromaDB, which made the collection more searchable in Claude Code. From there, Claude could identify notes that were not yet atomic and suggest categories across the larger collection.
The categorization has been especially valuable. These notes came from roughly ten years of different systems, so many of the connections were buried under old formats and inconsistent habits. I could have found some of them manually, eventually. But not enough to make the system feel alive again.
That does not mean the system is solved. It means the next bottleneck is clearer. I can retrieve more than I used to. I can identify notes that need work. I can see categories and connections that were previously hidden. Now the question is whether I can turn that into a steady output rhythm.
That is the real test.
If you suspect your PKM has become sophisticated procrastination, try a simple litmus test: in the last 30 days, what concrete output can you point to that directly came from your notes? Then ask the follow-up: what recurring ritual turns those notes into next actions, outlines, drafts, or decisions?
If you can’t answer both, you may not have a note-taking system yet. You may have a capture system with good branding.
The good news is that the fix probably is not another tool switch. Start smaller. Add one weekly review. Pick one project. Pull the relevant notes into a folder or outline. Produce one small artifact from them.
Then notice where the pipeline breaks.
That is where the next iteration starts.
Reply with the place your note system most often stops helping and starts becoming maintenance.