The Moment Everything Changed

Last Tuesday, in a parallel universe, a researcher in Tokyo published a paper about protein folding. She doesn't know it, but her paper contains the missing piece for a cancer treatment that a team in Berlin has been developing for three years. The Berlin team doesn't know it either. They'll never find her paper because they don't search for “protein folding”—they search for “oncology.”

Meanwhile, a grad student in Mumbai has been working on a manufacturing process that would make the Berlin treatment affordable. But he publishes in materials science journals. Different keywords. Different citations. Different world.

The cure exists. It exists in the space between these three papers. But it doesn't have a URL. This is the crime we commit every day: we let people die because information can't find itself.

The Beautiful, Terrible Truth

Every scientific paper published today changes the relationship between thousands of existing papers. Not metaphorically—literally. A discovery about CRISPR today retroactively transforms what a paper about manufacturing meant yesterday. A breakthrough in quantum computing this morning redefines what machine learning research from last year actually implies.

Information is alive. It grows, evolves, and restructures itself with every new piece added. But we've trapped it in amber—frozen documents with fixed addresses, connected by static links decided years ago by authors who couldn't possibly know what connections would matter today.

Try explaining this to the web: “The connection between this cancer research and that materials science paper didn't exist yesterday, but it exists today because of what was published this morning.” The web can't understand. URLs don't update themselves when reality changes. Links don't rewire when new knowledge makes old connections obsolete. Every search returns results based on relationships someone manually created in the past, not those that emerged one second ago.

Why Everything We've Built Is Wrong

We built a library when information wanted to be a conversation. We took human knowledge—the most dynamic, interconnected, living thing our species creates—and we made it pretend to be a book. We gave every idea an address, as if ideas live in one place. We connected them with links, as if relationships between ideas are permanent. We rank them by popularity, as if truth is democratic.

Even our newest solutions are just prettier prisons. Embeddings? They're fancier addresses—coordinates instead of URLs. We assign “dog” to [0.2, -0.5, 0.8, ...] and pretend we've captured its meaning. But what about “dog” in the context of evolution versus companionship versus allergies? What about how the meaning of “dog” changed when we discovered they can detect cancer by smell?

Information doesn't have a location. It doesn't even have coordinates. It exists in superposition—simultaneously connected and disconnected to everything else, with relationships that only crystallise when observed, shifting based on who observes, when they observe, why they observe, and everything discovered up to that millisecond.

We've been forcing information to be static in a dynamic universe. It's like insisting that rivers should stay in one place.

What AI Can See That We Can't

Imagine this: an AI discovers that a paper about butterfly wings is related to a paper about solar panels—but only when considering them alongside papers about crystalline structures, light diffraction, manufacturing processes, and biomimicry.

No human would ever search for those six topics together. The connection doesn't exist in any citation. No hyperlink connects them. But the pattern is there—a deep structural similarity in how both butterfly wings and certain solar cells manipulate light through nanoscale architectures.

AI doesn't browse from link to link. It can hold ten thousand papers in attention simultaneously and see patterns that only emerge when you consider papers A through Z together—patterns invisible when looking at any subset. It can see that the Tokyo protein research relates to the Berlin cancer treatment through a pattern that only appears when you also consider seventeen other papers from unrelated fields.

But we're still feeding AI through the tiny straw of URLs and keywords, as if it needs to navigate the way humans do. It's like giving a telescope to someone and insisting they only look through a keyhole.

The Question That Haunts Me

I don't know how to fix this. I don't know how to build a system where information finds itself. But I know we need something radically different—not better search, not smarter links, but a complete reimagining of how information exists and relates.

Maybe it looks like this: instead of searching, you describe a problem, and the system recognises which constellation of knowledge across all human work could solve it—even if those pieces have never been explicitly connected. Instead of URLs pointing to fixed documents, information exists in a fluid state, reorganising itself around each query based on everything known up to that moment.

The web was miraculous for its time. It connected documents when documents were all computers could handle. But we don't need connected documents anymore. We need living knowledge—information that reorganises itself with every new discovery, that finds its own patterns, that recognises relationships we haven't manually created.

What I'm Not Saying

I'm not saying I have the architecture. I'm not saying embeddings are the solution. I'm not saying any current approach is sufficient. I'm saying that we're living through a crime against human knowledge. We have the tools—AI that can process meaning directly, transformers that understand context, compute power beyond imagination—but we're using them to navigate a map drawn for a different species in a different century.

The Choice We Face

Every morning, we wake up and pretend the web makes sense. We type URLs as if information should have addresses. We create links as if we can predict what connections will matter tomorrow. We search as if the answer we need was written down somewhere, rather than existing in the space between everything that's been written.

Or we admit what we secretly know: the web is already dead. It died the moment AI could see patterns we can't manually link, find connections without URLs, and recognise relationships across domains that have never spoken.

What comes next will be built by someone who stops trying to improve search and starts building recognition. Someone who stops organising information and starts letting information organise itself. Someone who realises that in a world where AI can hold all human knowledge in attention simultaneously, making it crawl from link to link is absurd.

The Real Question

I started by telling you about three researchers whose work could save lives if it could find itself. Here's what keeps me up at night: how many cures are we missing? How many breakthroughs sit in plain sight, separated only by the artificial boundaries of disciplines, keywords, and URLs? How many problems have already been solved by someone who didn't know they solved them?

The Tokyo researcher, the Berlin team, the Mumbai student—they represent millions of connections that should exist but don't. Every day, someone dies from a disease whose cure exists in the space between unconnected papers. Every day, someone reinvents something already solved in a field they'll never search.

I don't know how to build what comes after the web. But I know we have to. Because somewhere right now, a researcher is publishing a paper that changes everything—and no one will know for years, decades, or ever. Not because the information isn't there, but because information can't find itself. That's the crime. That's the revolution. That's the question I can't stop asking: What if information could recognise itself?

Someone reading this knows how to solve it. Maybe it's you. Maybe it's the person you'll forward this to. Maybe it's someone who hasn't realised they're thinking about this problem yet. But somewhere, someone just realised that information doesn't need to be searched—it needs to be understood. And understanding doesn't happen in documents or links or even embeddings. It happens in the space between everything we know, in patterns that only exist when all knowledge can see itself at once. The web was the best we could do with the technology we had. It's not the best we can do with the technology we have.