This is a filing cabinet. Notice the folders labeled A through Z?

So... Where do you file that great idea you had about combining Ghengis Kahn's tactical strategy with a Warhammer style dystopian fiction? Under "G?" How about "Scribbles?" Or maybe you create a whole new folder called "Random Ideas" that you'll never remember to check.

This is essentially what we're doing with computers right now. We're using organizational systems designed for a world that no longer exists, trying to cram the complexity of human thought into rigid hierarchies that made sense when information was scarce and simple.

Here's where I drop AI again. This one is from Marc Andreesen, who annoyingly makes a lot of sense here.

Artificial intelligence isn't just changing what computers can do—it's forcing us to completely reimagine what computers are and how we should think about them.

Let's just say that humans and AI think in fundamentally different ways. (I know AI doesn't "think," but for argument let's go with it.)

First you.

When you try to recall where you put your keys, you don't search through an alphabetical mental index. You think about what you were doing, who you were with, what you were feeling. Your brain creates connections—"I was stressed about the exam, I remember putting them down while I was on the phone with Mom, probably near the kitchen because I was making coffee." Your memory is a web of associations, not a filing system.

Now, look at your computer.

Files in folders, arranged in hierarchies. Documents organized by type or date or project. Perhaps you have a desktop folder that says "AprilJunk24." (I, sadly, sure do.) It's the digital equivalent of that filing cabinet I hated, and it's becoming increasingly absurd as our lives become more complex and interconnected.

Turns out, AI doesn't "think" in folders either.

When an AI system processes information, it creates vast networks of connections, understanding concepts not by where they're filed but by how they relate to everything else. It's more like how you actually think, but exponentially more complex.

We're trying to collaborate with AI systems using organizational principles that neither humans nor AI naturally use. It's like trying to have a conversation by passing notes through a bureaucratic office where every thought has to be properly categorized and filed before it can be shared.

When you query my Claude, it's not searching through a series of remote filing cabinets in Dario Amodei's office. Instead, it's accessing the encoding information from embedded "vectors."

WTF is a Vector, Victor?

A vector is a string of numbers that represents a point "floating" in a humanly-unimaginable, massive, multidimensional space.

Think of it like GPS coordinates, but instead of just latitude and longitude (2 dimensions), you have thousands of coordinates that capture:

• meaning, context, emotion, and relationships.

If you had to describe the location of everything in your room, you might use three dimensions—x, y, and z coordinates. But what if you wanted to describe not just where something is, but what it means, how it relates to other things, what emotions it evokes, what memories it triggers? You'd need way more dimensions. AI systems use spaces with 1000s upon 1000s of dimensions. We literally run out of letters in the alphabet to describe all these dimensions.

In this crazy mathematical space, similar ideas naturally cluster together. "Dog" and "puppy" end up close to each other, but so do "loyalty" and "friendship." The computer doesn't need someone to manually create a "pets" folder—it discovers that dogs and cats are related by seeing how they occupy similar regions in this multidimensional space.

This is fundamentally different from how we've been organizing information. I found this (now classic) video from Google on "High Dimensional Space" to be useful.

Instead of forcing everything into predetermined categories, we let the computer discover natural relationships. Your essay about urban planning might end up connected to your economics notes not because you put them in the same folder, but because they occupy nearby regions in this conceptual space.

As I get better with working this way, I find that this doesn't just store these connections; it helps you discover new ones, surfacing unexpected relationships. This concept is often referred to as "discoverability."

I see glimmers of this kind of thinking in tools like Obsidian and Roam Research. The internet is filled with knowledge managers designing their "second brain." The real revolution happens when this quirky networked connection idea becomes the default way computers work, not just specialty tools for knowledge workers.

If we're not organizing by topic, what do we organize by?

I've been experimenting with this, and I'm finding that the most useful way to #tag things isn't by what they are, but by where they are in my thinking process.

Instead of labeling something "#economics" or "#art project," I tag it as "#improv," "#rough," #snippet or "#essay." When you're looking for inspiration, you don't want to dig through your finished essays. You want to browse those half-formed thoughts, those wild early hunches that might spark something new. Geghis Kahn's squadron command strategy ideas, and the imperial insignia protocol from Warhammer can now be mixed and matched together like alchemy. New strands of "idea DNA" that can be graphed together.

This shift feels unsettling to me.

We're moving away from asking "What is this?" toward asking "What stage is this in?" and "What's it ready to become?" Your ideas become less like static files and more like living things with their own life cycles.

I'm beginning to think we will completely transform how we think about expertise and careers. If our computers organize around production stages rather than subject matter, we stop being "the economics person" or "the art person." Instead, we become people who are good at "taking early ideas and developing them," or people who excel at "connecting disparate concepts," or people who can take "rough prototypes and polish them" into something shareable.

Maybe we're heading toward a world where we're all generalists who can dive deep into any domain, moving fluidly between projects and roles as needed? The networked computer doesn't care if you're working on sustainable urban planning on Monday and a poetry project on Wednesday. It just cares about helping you move each idea forward, whatever stage it's in.

Your value isn't in what you know—it's in how you process, connect, and develop ideas. The computer becomes your thinking partner, not your filing system.

In the old model, you had an idea, you developed it privately, you turned it into a product, and you sold it. Value was created through ownership and scarcity.

But in a network-based system, value is created through connection and abundance. Your half-formed idea becomes valuable not because you own it, but because it connects to someone else's complementary insight in a way that creates something neither of you could have achieved alone.

I've often pointed out that open-source software development already works this way. The most innovative companies are starting to work this way internally, breaking down traditional departmental silos in favor of cross-functional networks where ideas and expertise flow freely.

But imagine if our computers were designed from the ground up to support this kind of collaboration. Instead of files and folders, you'd have nodes and connections. Instead of working on isolated documents, you'd be contributing to living networks of ideas that grow and evolve as more people connect to them.

This doesn't mean becoming more like machines. It means becoming more like ourselves—more connected, more creative, more collaborative. The rigid hierarchies and artificial categories that we've been using to organize information were never natural to human thinking anyway. They were compromises we made when we had to work with limited, static tools.

It's possible we could build computers that work the way we think, that support the way we actually create and connect and grow.

Death to filing cabinets.

We'll see you next time.

If you vibe to the ideas I express here, please consider subscribing and sharing with friends. Thanks for reading.

Nye Warburton is a system designer from Savannah, GA. This essay was improvised and iterated over several walks to work with Otter.ai and alchemically refined with Anthropic Claude.

For more information visit: https://nyewarburton.com

• WTF are Agents? - June 1, 2025

• Bear_Rabbit_Robot: Mapping AI's Latent Space - March 13, 2025

• Rethinking Talent for Generative Game Pipelines - December 23, 2023