CS PAPERS DEEP-READ · PAPER 35

As We May Think

Vannevar Bush · The Atlantic Monthly · 1945

中文 →

What did this essay do?

Every time you click a blue link and jump to related content, or bookmark a string of useful pages into a folder, you're using an idea — "click and jump to the related thing" — that an American scientist named Vannevar Bush had already thought through in 1945. He wrote it in an essay called As We May Think. There were no personal computers then, and no internet. He imagined a desk called the Memex — a "memory-extension machine."

First, a puzzle

World War II had just ended and human knowledge was exploding — more books, taller stacks of papers. But Bush noticed something maddening: the more there is, the harder it gets to find the one thing you need. Worse, plenty of important findings were published and then buried in the flood, never dug up again. The problem wasn't that we produced too little — it was that storing things is easy, finding them is hard, and connecting the related ones is harder still. Our way of finding information had barely changed in centuries.

His idea

First, said Bush, solve "hold it all." Using the era's cutting-edge microfilm (photographing pages down to tiny images), he predicted the whole Encyclopædia Britannica could be shrunk to the size of a matchbox. So the Memex desk could, in principle, hold every book you'll ever read and every note you'll ever take, called up to a screen at a touch. But that wasn't the real breakthrough — the real breakthrough was how to connect them all.

How did he reach the key step?

Bush stared at how libraries file things and spotted the flaw: books are shelved alphabetically or by class number, layer by layer (big class, then sub-class, then sub-sub-class). Finding one means climbing down a tree — and a given item can only live in one slot. But, he argued, that's not how the human mind works — the mind works by association: think of one thing and it snaps instantly to a related one.

So why can't a machine do the same? Here is the Memex's soul: let the user tie any two items together — and afterward, looking at item A, press a button and the machine instantly pulls up the B you tied to it. Tie a string of related items one after another and you've woven a "trail": you can name it, save it, walk the whole thing again later, even copy the entire trail to a friend. This is the ancestor of today's hyperlinks, bookmark folders, and even knowledge graphs.

What it led to

The people who later invented the mouse and built the first hypertext systems (Douglas Engelbart, Ted Nelson) openly said this essay inspired them. That "click and jump to the related thing" bloodline grew up and became the World Wide Web we use every day. An essay with not a single formula — not even one machine ever built — sketched the shape of the whole information world half a century early.

Remember one thing

Back in 1945 Bush saw it: rather than shelving knowledge rigidly by class the way a library does — one item, one slot — let people do what the brain does, tying related items together on the fly into named, shareable "trails." That's the source of the hyperlink and the Web. (Though the Memex was mechanical microfilm and was never actually built, and those "trails" still had to be tied by hand, one at a time — what really made vast information findable later was the automatic search engine, not hand-made trails.)

Want the Memex diagram, how the associative trail works, and what it actually predicted? → switch to the deep read