Your Notes Are Full of Gavagais
Your notes aren't disorganized. They're full of unresolvable references — written for a reader who shares context that no longer exists.
There's a note I wrote in late 2023 that says: "Follow up with him about the timeline — he was worried about the equity structure." I know which project it was. I still don't know who "him" is.
That's not a memory problem. That's a reference problem.
W.V.O. Quine introduced the sharpest version of this in his 1960 book Word and Object, through a thought experiment he called "gavagai." A linguist is standing with a native speaker when a rabbit runs by. The native says: "Gavagai!" The linguist notes it down as probably meaning "rabbit." But Quine points out the problem: it might equally mean "undetached rabbit parts," or "a momentary rabbit-stage," or "rabbithood," or "food." All of these are present whenever a rabbit appears. The behavioral evidence is identical in every case. There's no way to know, from what's observable, which thing "gavagai" actually points to.
Quine called this the inscrutability of reference. What a word picks out only resolves against a background of shared assumptions — an ontological framework both speakers bring without noticing. Lose that framework, and the reference floats free.
What notes are actually doing
Most notes get written in a state of maximal shared context. You wrote "follow up with David about the equity concern" immediately after the meeting, with the conversation still alive in your head, with David's face and company and specific worry all fresh. At that moment, "David" couldn't be anyone else, and "the equity concern" could only mean one thing. The reference felt obvious because you had the background. You were your own native speaker.
But you are not that reader anymore. The project finished, the people moved, the vocabulary drifted. Now you're the linguist standing in an empty field, holding a word that could point to anything.
This gets compounded in a specific way when you add AI to the picture. A 2025 paper on retrieval-augmented generation found that more than 60% of follow-up queries contain unresolved coreferences — pronouns, demonstratives, references that can't be resolved without the conversational context that wasn't stored alongside them. Feed an AI assistant your personal notes and ask it to find "what I learned from that conversation with the nervous investor," and it will look for "nervous," "investor," "conversation," and whatever else you gave it. It will find dozens of candidates. None will clearly be the right one. The retrieval isn't broken. The references just weren't written to be resolved.
Binding the reference at capture time
After the fact, there's no fixing this. The context is gone. You can't recover who "him" was when there's no longer any shared world to triangulate against. Quine's linguist had at least that much: a rabbit still running somewhere.
The fix has to happen when you write. Not writing more prose — just committing, at the moment of capture, to what the reference actually points to. Tying the pointer to a specific, structured thing that doesn't depend on surrounding context to be understood.
This is what person records and typed entities in a knowledge base actually do. Not the ambiguous "him" of a note, but a link to a specific person: David Kwan, VP of Product at Meridian, first met October 2023, introduced through Sarah at Acme. Not "the partnership conversation," but a note attached to a meeting record with a date, attendees, and the specific concern flagged. The reference becomes fixed. The entity doesn't decay when the context does.
Quine's point was that a word without a background theory can't be interpreted. Typed entities give that background theory somewhere to live: outside the note, in a structure that persists beyond the moment of writing.
The cost of leaving things vague
None of this is obvious while you're writing. The note feels complete. You understood it perfectly, the same way Quine's native understood "gavagai" perfectly. The problem only surfaces later, when the context is gone and the reader is a stranger — or is you, two years out, going through a knowledge base that has slowly become a collection of pointers with no targets.
The question isn't how to write more complete notes. It's how to make the references in your notes resolvable by default, without adding friction. Structured capture, explicit entity links, person records: not organizational tidiness, but the architectural difference between a note that points somewhere and a note that might point anywhere.
Quine spent a book worrying about whether we can ever know what another speaker means. He was right to. But in personal knowledge, the problem is simpler and stranger: you need to know what you meant. And that's only possible if, at the moment of writing, you committed the reference to something more durable than context.
The rabbit won't be there when you come back.
Asgeir Albretsen is the founder of Harbor.