Vannevar Bush wrote “As We May Think” in 1945, just as the war was ending, and got almost everything right. The problem he saw was this: the amount of human knowledge was expanding faster than our ability to use it. Not store it — use it. The bottleneck was access, and access as it existed then was alphabetical, taxonomic, artificially sequential. Indexes led to subindexes led to the shelf. The encyclopedia had solved nothing. It had just organised the filing cabinet.
What Bush proposed instead was a device he called the memex. A desk-sized machine with microfilm reels and projection screens and a keyboard. Visually, it looks quaint now — more telegraph office than internet terminal. But the mechanism Bush described was not a filing system. It was an associative trail.
The idea was simple and still radical. When two items were connected in the mind — this article to that article, this formula to that result — the connection itself became an artifact. You could name it. Save it. Walk back down it. Other users could follow trails you had left, or branch off from them, or add their own annotations. The machine would not organise knowledge the way a library does, from outside in, by category and genus and species. It would organise it the way the mind does: through association, through the link made in the moment of thinking.
Bush had identified something that nobody in the information sciences had quite named before: the mind is not a sequence machine. It is a network machine. It thinks in trails, not tables of contents.
The reason this matters philosophically rather than just technologically is that Bush was making an implicit claim about the nature of knowledge. If the mind works associatively, and if our systems for storing and retrieving knowledge work sequentially, then our systems are not neutral. They are a distortion. Every library, every encyclopedia, every index imposes a false linearity on a fundamentally nonlinear process. The knowledge we cannot follow the trail to is knowledge we effectively do not have.
The person who took Bush’s idea and made it into a religion was Ted Nelson, and what Nelson built was not the web. That is the crucial thing to understand about Nelson: his vision was larger, stranger, and more philosophically serious than what Tim Berners-Lee eventually shipped.
Nelson coined the word “hypertext” in 1963. He spent the next several decades trying to build Xanadu, a system so ambitious and so precisely conceived that it remains, in certain important respects, unrealised. The web, to Nelson, is a diminishment. He has said so repeatedly, in increasingly exasperated tones, and he is right.
The web’s links are one-directional. You can link from page A to page B, but page B does not know it has been linked. The original document has no record of what points to it unless a search engine has indexed the fact. This means that when you quote something on the web, the thing you quote does not know it has been quoted. The connection exists in one direction only, which means it is not really a connection. It is a pointer. A ghost of the link Bush imagined.
Xanadu had transclusion instead. A quoted passage would remain part of its original document — always, inextricably — and the quoting document would know what it was borrowing. Every link would be two-way. Every connection would be visible from both ends. The authorship, the provenance, the context would travel with the text wherever it went.
The philosophical consequence of transclusion is large. A web built on it would be a web where nothing could be taken out of context, because the context always came with it. It would be a web that remembered. The web we got is a web that forgets — pages vanish, links rot, the document that was quoted in ten thousand places disappears and the quotes drift free of their origins, untethered, easily weaponised. The associative trail that Bush imagined depends on the trail staying intact. The web keeps losing sections of itself.
Nelson is cranky about all of this. He has been cranky for fifty years. He is not wrong. What he understood, and what the architects of the actual web underweighted, is that the link is not decoration. The link is the unit of meaning. Build a system where links are fragile and one-directional and you have not built what Bush described. You have built a system that mimics the associative trail without honouring what makes the trail meaningful: the connection held in both directions, the memory of where the text came from.
What is stranger and more interesting is that Jorge Luis Borges was doing all of this in fiction before either Bush or Nelson had coined a term for it, without a machine, and with a precision that neither of them quite matched.
Borges was a librarian. This is not incidental. He spent years in the Municipal Library of Buenos Aires cataloguing books, living inside the structure of recorded knowledge — its edges, its gaps, its loops, the way certain texts silently organise the ones around them without anyone having planned it that way. That experience infected his thinking permanently.
The Library of Babel, published in 1941, four years before Bush’s essay, is not a story about books. It is a theory of information. An infinite library containing every possible book, every possible arrangement of characters, which means it contains every true statement and every false one, every masterpiece and every incoherent string. The problem is not the absence of knowledge but the impossibility of locating it. The library contains the answer to every question ever asked. The answer exists. The trail to it does not. Borges had articulated, in the form of a nightmare, the exact problem Bush was trying to solve.
The Garden of Forking Paths, published the same year, goes further. It is a spy story on the surface. Underneath it is a meditation on time as a branching structure rather than a line. The garden in the story is a novel written by the narrator’s ancestor — a novel that contains all possible outcomes simultaneously, in which every fork is taken, in which nothing is resolved by elimination. The story you are in is not the true story. It is one path through a space that contains all paths.
This is not a metaphor for hypertext. It is hypertext, stated as philosophy, before the word existed. Borges understood something about the nature of knowledge and narrative that the information scientists were still working toward: the sequence is not the thing. The sequence is one path chosen out of the space of all paths. Change the path and you change the meaning. The meaning was never in the sequence. It was in the space.
Borges was also, notably, working in traditions he knew consciously. He understood Arabic literature, the Thousand and One Nights, the way stories can nest inside stories inside stories until the frame disappears. He understood the Talmud, which is a document structured as commentary on commentary on commentary, where every margin is full of argument and the argument never resolves. He understood Dante, who built a cosmology out of the idea that every soul in the afterlife is an interpretive key to every other. These are all, in different ways, hypertext architectures — networks of meaning in which navigation through the connections is the act of understanding, not a path toward it. Borges just saw them that way explicitly, and said so in prose that is itself structured more like a network than an argument.
The short story as Borges wrote it is not a container for a plot. It is a node in a larger network of texts, ideas, references, and traditions that the story activates by linking to them. You read a Borges story differently if you have read the previous one, differently again if you have read the source texts he is quietly rewriting. The meaning is not in the text. It is in the traversal.
The blog, when it arrived, looked like it might finally build the memex in public.
The reverse-chronological structure was a convention, not a necessity — what mattered was the link. The blogosphere at its peak, roughly 2002 to 2008, functioned as something close to what Bush had imagined: a network of trails made by minds thinking in public, cross-linked, annotated, responsive. You wrote something, I linked to it, someone else linked to both of us, and the conversation existed in the links as much as in the posts. Technorati tracked the connections. Blogrolls were explicit declarations of intellectual neighbourhood. When a post caught fire in that network, you could follow the chain of references backward through the web’s actual intellectual topology. The trail was visible.
It didn’t last, partly because the tools didn’t support it and partly because most people didn’t want it. The associative mind is not the average mind. Given a publishing platform, the average impulse is toward audience, not network. The blog became a broadcast medium with a comments section, which is the inversion of what the hypertext tradition had in mind. Twitter accelerated this — shorter, faster, more reactive, less structural. The trail got replaced by the feed. The feed has no memory of where anything came from.
But the blog hasn’t died, and the ones that survived the platform collapse are the ones that kept the associative logic intact. A writer who follows a thought across three hundred posts over fifteen years, linking forward and back, building a network in which each individual piece is more meaningful in context than it is in isolation — that writer is building something closer to what Bush imagined than anything a social platform has managed. The associative trail is still there. It requires intention to maintain, and the web makes that harder than it should be, but the trail can hold.
The digital garden movement is the most recent attempt to recover this logic — a kind of gardened hypertext, notes and fragments and essays linked to each other by association rather than by date, the temporal sequence replaced by the topological one. It is less polished than the blog and more honest. It looks more like how the mind actually stores things: imperfect, cross-referenced, provisional, alive.
There is a philosophical claim underneath all of this that Bush intuited, Nelson made explicit, Borges demonstrated fictionally, and the blog keeps testing against reality.
The claim is this: meaning is not in the document. It is in the relationship between documents.
Print culture thinks in sequences. A book has a beginning, a middle, an end. The argument develops. The conclusion resolves. Meaning accumulates directionally. When you reach the last page, you have the thing. This model of knowledge is also, not coincidentally, the model we use for almost everything else: the lecture, the lesson, the career, the life. Beginning, development, resolution.
Hypertext as epistemology says this is wrong, or at least partial. Meaning is not accumulated directionally. It is generated navigationally. It depends on where you came from before you arrived at this page. It depends on where you go when you leave. The same node in a network means different things depending on the path that brought you to it.
This is not relativism. The distinction matters. Relativism says there is no truth, only perspectives. Hypertext epistemology says there is truth, but it is path-dependent. The map of the territory changes based on how you enter it. The meaning of a document is partly constituted by the documents it links to and the documents that link to it. Strip a page from its network and you haven’t preserved the meaning. You have preserved the text. The meaning was in the connections.
The consequence for writing is significant. If you accept this, then the most important thing about any piece of writing is not its internal argument but its external relationships. Where does it point? What points to it? What becomes possible after you have read it that was not possible before — not because you have the information it contains, but because it has opened a path to the next thing? The hypertext mind writes with this in mind. The text is a node. The node gets its significance from the network.
Borges knew this. The story in the garden of forking paths doesn’t mean the same thing until you choose a path through it, and it means something different if you take the path a second time. Nelson knew it: the transclusion he wanted would have kept every text tied to its origins because the origin was part of the meaning. Bush knew it, which is why he focused on the trail rather than the document — the trail was the unit of knowledge, not the file at the end of it.
The blog is the closest approximation we have built, in public, to this model. A post doesn’t live alone. It lives in a network of other posts, linked and linking, contextualised by everything that came before it and everything it points toward. A reader who arrives at post 347 having read posts 1 through 346 is reading a different post from the reader who arrives from a search engine with no prior context. The text is identical. The meaning is not.
This is what Vannevar Bush was imagining in 1945, writing in a world where information was expanding faster than the trails through it. He wanted a machine that would let the mind move the way the mind actually moves: not from A to B to C but from A to the thing A reminds you of, to the thing that connects the two, to the question that opens up in the gap between them, to the document you saved last Tuesday that turns out to be the key.
The machine arrived as the web. It got the links right and almost everything else wrong. Nelson is still waiting for his transclusion. The garden keeps forking. The trails we make are the best record we have of how we actually think — associative, lateral, alive to the connection that arrives before it can be justified.
What the web gave us, even in its diminished form, is the public associative trail. Once that existed, the possibility changed. The blog, the wiki, the digital garden with its backlinks — these are all attempts to honour what Bush named and Borges already knew: the mind does not think in chapters. It thinks in links. The most honest thing you can do with that fact is let the links show.
Network
Series: Hypertext and Mind
- The Trail the Mind Makes — this essay
- Vannevar Bush and the memex (depth node to write)
- Ted Nelson and the Xanadu dream (depth node to write)
- Borges as hypertext thinker (depth node to write)
- The blog as associative mind (depth node to write)
- Hypertext epistemology (depth node to write)
Sparked by
- Meaning Through Navigation, Not Sequence — the fragment that names the core claim: meaning emerges through navigation rather than sequence
Related essays
- The Text-Based Ontologist — the seed phrase behind the wider practice of writing as ontological work
- The Spirit of the Textual Underground — the linked web as home territory for the associative mind
- The Entanglement Dialogue (to locate or write)