What Feynman Kept in His Head
Every note-taking tool is built to capture what you know. Richard Feynman kept a list of what he didn't know yet — and treated it as his primary intellectual tool.
The mathematician Gian-Carlo Rota, who worked alongside Richard Feynman at Caltech, described how Feynman prepared for discovery: "You have to keep a dozen of your favorite problems constantly present in your mind, although by and large they will lay in a dormant state. Every time you hear or read a new trick or a new result, test it against each of your twelve problems to see whether it helps."
Not twelve thousand notes. Twelve questions.
What Feynman kept wasn't a record of what he knew. It was a list of what he didn't know yet: open problems that functioned as a filter for everything he encountered. New information got tested against the list. If something illuminated one of the twelve, even slightly, it became interesting. If it didn't, it was someone else's problem.
This runs exactly backward from how almost every note-taking tool works.
The hidden assumption in every note app
Every knowledge management system ever made treats the note as the unit of knowledge. You read something, you write it down. You finish a meeting, you capture what was decided. The whole apparatus — tagging, linking, searching, organizing into folders — is designed to hold and retrieve conclusions.
The hidden assumption is that knowledge is settled. You know it or you don't. You captured it or you lost it.
But a question doesn't work that way. A question isn't something you possess. It's something you're inside. It doesn't sit quietly in a database waiting to be retrieved. It operates on incoming information whether you invite it to or not. That's what Feynman understood: the question is generative in a way the answer never is. An answer closes. A question stays open.
Warren Berger spent years studying how breakthroughs happen and published A More Beautiful Question in 2014. His argument across dozens of case studies is that the most productive innovations didn't begin with better information. They began with someone asking a question nobody else was asking. The question preceded the knowledge and generated it.
Luhmann's communication partner
Niklas Luhmann, the German sociologist who spent forty years building a 90,000-card Zettelkasten, described his system not as a filing cabinet but as a "communication partner." He didn't want retrieval. He wanted surprise.
The goal was to be put somewhere unexpected. A card would connect to another card in a way he hadn't anticipated, and Luhmann would find himself following a thread he hadn't planned. The system was designed to generate new questions, not confirm old answers.
Most digital tools inspired by Luhmann's method miss this entirely. They give you the links. What they don't give you is resistance — the friction of encountering something you didn't expect. You can traverse a modern knowledge graph for hours and never be surprised, because every connection is one you made deliberately.
Feynman's twelve problems created a different kind of system. Because the questions stayed open, new information had something to bump against. A result from a completely unrelated field might illuminate one of the twelve. The questions were the connective tissue; everything else was potential signal.
What you can't export
Export your notes from any app and look at what you get. You'll find documents, highlights, summaries, action items. What you won't find is what you were wondering about.
The questions that drove a project — the things you were unsure of, the tensions you hadn't resolved, the hypotheses you were testing — almost never make it into notes. They lived in your head while you worked and dissolved when the project closed or the task got marked done. This is a real loss, not because the questions were precious objects, but because they carry context that answers don't. An answer without its question is a fact. A fact with its question is a lens.
When you return to a project after three weeks away, you don't just need a summary of what was decided. You need to know what was still uncertain, what was being tested, what hadn't been figured out yet. A good question is a map back into the problem. Most notes apps don't have a place to put it.
The question as a primitive
I started keeping a short list of open problems after reading about Feynman's technique. Eight questions, roughly, that I update when something gets resolved or a new one surfaces. It's the only part of my knowledge base I look at every week.
What surprised me was the discomfort. Unlike a note, a question doesn't give you the satisfaction of having captured something. It sits there, unresolved, asking for attention. That discomfort turns out to be the point.
Structured knowledge helps here in a way raw text doesn't. A project record is already asking "is this finished?" A person record is asking "what do I still not know about this person?" The entities themselves are question-shaped — incomplete by design, accumulating what they don't yet have.
Feynman's twelve problems weren't stored anywhere. They lived in his head, because that's where questions live. But the structure around them — what he read, who he talked to, what he chose to test — that was the system that made them generative.
Most notes apps have the structure. They're still missing the questions.
Asgeir Albretsen is the founder of Harbor.