The original Memex
In July 1945, as the Second World War ended, Vannevar Bush published an essay in The Atlantic called "As We May Think." Bush had directed American wartime science. He saw that the deluge of new knowledge was already overwhelming the people who needed it, and argued that the next great human project was to make that knowledge accessible again.
The device he proposed was called the memex: memory plus index. A desk, wired with microfilm reels, a translucent screen, a keyboard, and a stylus. Researchers would store everything they read. They could photograph new pages onto the film. They could write notes in the margins. And, crucially, they could link documents to each other by association, the way their own minds did, so that one thought could lead them to the next without having to remember where it was filed.
Associative trails
Bush called these connections associative trails. It was the part of the idea he cared about most. Existing indexes, he wrote, worked alphabetically or by number. Rigid, artificial paths that didn't match how humans actually thought. The mind worked by association. One fact suggested the next. The next triggered a memory. That memory opened a new question.
The user could build a trail of his interest through the maze of materials available to him. Vannevar Bush, As We May Think, 1945
A scientist could follow a chain of reasoning through their library without losing their place. They could share a trail with a colleague. They could inherit one from a teacher. Trails were knowledge that travelled.
Everyone who mattered read it.
For decades the memex existed only on paper. But the essay landed in the right hands.
Douglas Engelbart read Bush as a young engineer and, by his own account, "became infected with the idea." He demonstrated the first graphical computer interface in 1968 (the mouse, video conferencing, shared editing) in what became known as The Mother of All Demos. He was building Bush's device, updated for the computer age.
Ted Nelson read Bush and coined the word hypertext, which he defined as "non-sequential writing with reader-controlled links." The word itself is a paraphrase of the memex.
Tim Berners-Lee proposed the World Wide Web. He later acknowledged that he had independently arrived at a system Bush had already imagined.
Every act of linking on the internet traces back, at some remove, to this essay.
Memex animation: Bush's original diagrams brought to life. Watch on YouTube.
The modern incarnation
Bush's problem was information overload. Ours is one he didn't live to see: AI coding agents that can write software faster than anyone can direct them, pulling from oceans of code, documents, and prior decisions that no single human mind can hold.
He wrote in 1967 that the real challenge was never raw speed. It was "selective access" to vast knowledge stores. That's exactly the problem an AI-native software team runs into today. The agents are fast. The humans can't keep up. Decisions scatter. Context rots. Teams end up rebuilding work that was already done, or building the wrong thing because nobody, human or machine, can find what was decided and why.
Memex AI is our answer. A shared graph of strategy, decisions, and blueprints that every agent and every engineer on the team can reason over. Trails of connected thought, queryable by both sides. Knowledge that travels.
Honouring the vision.
We didn't borrow the name lightly. Calling this tool Memex is an acknowledgement: that the pattern we're building has been waiting since 1945, and that the deepest problem Bush named hasn't been solved. Only reshaped.
A new relationship between thinking man and the sum of our knowledge. Vannevar Bush, As We May Think, 1945
That was the goal Bush set. It's still the goal.