The Memex: The 1945 Idea Behind Your Second Brain

In 1945, an engineer named Vannevar Bush described a machine he called the memex: a personal device where you’d store all your books, records, and notes, and — the important part — link them together by associative trails you built yourself, with the connections between documents being as valuable as the documents. He was describing, decades before the technology existed, almost exactly the linked-notes “second brain” that a lot of us keep today. It’s worth knowing that lineage, because it clarifies what these tools are actually for, and it points straight at the one piece Bush never solved.

What Bush actually proposed

The memex wasn’t a search engine or a filing cabinet. Bush’s key insight was that the human mind works by association — one thought leads to the next by connection, not by alphabetical index — and that a knowledge tool should work the same way. So the memex let you build trails: link this document to that one because they relate in your thinking, then follow the trail later. The organizing principle wasn’t a rigid category tree; it was the web of connections you personally drew between things. Two ideas mattered most, and both are quietly radical:

• The links are the value. Not just the documents — the trails between them. A fact connected to ten others is worth more than the same fact sitting alone, because the connections are where understanding lives.
• It’s personal and curated. The memex was yours, reflecting how you think, not a public index built for everyone.

The web went a different way

You’d think the web fulfilled this — it’s hyperlinked documents, after all. But the web became something else: public, vast, optimized for reaching everyone, and mostly not curated by the individual using it. The memex was private, personal, and curated by you. What actually inherits Bush’s vision isn’t the web — it’s the linked personal knowledge tools: the wikis, the [[wikilink]]-based note apps, the “second brain” systems where you connect your notes into your graph. Those are the memex, finally buildable. The trails Bush drew by hand are the wikilinks you type, and the graph they form is the associative web he was after — which is the whole reason I write everything down and link it instead of hoarding disconnected files.

The web gave everyone every document. The memex wanted to give you the connections between your documents. Those turned out to be very different projects.

The problem Bush couldn’t solve

Here’s the part that makes this more than a history lesson. Bush’s vision had a fatal practical flaw, and everyone who’s built a personal knowledge system has felt it: who maintains the trails? Building and keeping all those associative links is work — constant, tedious, never-finished work. Link this to that, notice the connection over here, keep it current as things change. It’s exactly why most “second brains” die: not because capturing is hard, but because the ongoing connecting and tending is a burden people can’t sustain. The memex assumed a tireless curator that human attention can’t actually supply. The graph decays into disconnected notes, and the dream quietly dies in a folder nobody opens.

What’s finally solving it

The reason this is worth revisiting now is that the maintenance problem — the thing that killed the memex dream for 80 years — is exactly what today’s tooling can finally take off your hands. An LLM can do the tireless bookkeeping Bush’s vision required: read a new note, find where it connects to what you already have, draw the links, keep the index current, flag contradictions and orphans. The human supplies the judgment and the sources; the machine maintains the trails. That’s the leap — not “capture,” which was always easy, but connection at scale without a human doing every link by hand. It’s the premise behind building a second brain that files itself, and it’s a big part of why you probably don’t need a vector database to have a genuinely useful personal knowledge store — a maintained web of linked notes is a different, often better thing than a pile of embeddings.

Why the lineage matters

Knowing the memex changes how you hold your own notes. It tells you the point isn’t accumulation — it’s connection, because Bush was right that the trails are where the value is. It tells you the hard part was always maintenance, so if your system is decaying, that’s the expected failure, not a personal one — and the fix is to offload the tending, not to try harder. And it tells you that the thing you’re building in a linked-notes app isn’t a fad; it’s an 80-year-old idea whose enabling technology finally showed up. We’re not inventing the second brain. We’re finally building the memex — and, for the first time, we’ve got something willing to maintain the trails. If you’ve built a personal knowledge web that actually holds together, I’d like to hear how you tend it.