You have sat in this meeting. A room of capable people, each fluent in their corner of the work, an hour of real discussion, and then everyone stands and no two of them could give you the same account of what was decided. Each of them understood the problem. The intelligence the work demanded sat in the room the whole time. The meaning between them came apart, the shared understanding that was supposed to hold across the conversation and walk out the door.

That failure has a name: a shortage of coherence, the capacity of meaning to keep its shape as it moves across people, across scale, across time. Intelligence is usually abundant. Coherence is what a room runs out of.

The same failure runs through the history of ideas, and it scales. Two of the eighteenth century's finest chemists studied the same gas and meant different things by it: Joseph Priestley saw dephlogisticated air where Antoine Lavoisier saw oxygen, each reading the same experiment through a framework that gave it a different sense.1 Widen the frame and the words themselves come loose: in 1996 a physicist slipped a paper of deliberate nonsense, dressed in the right vocabulary, past the editors of a respected cultural-studies journal, because in that corner of the conversation the marks of meaning had drifted free of meaning.2 Widen it once more, across time, and a whole literature gives way: when researchers set out to reproduce a field's published results, fewer than half held, a structure of competent, peer-reviewed work that came apart on contact with retesting.3 Each of these is a coherence failure, the meaning of the work failing to hold across two minds, across a field, across a generation.

Now consider what you reach for to hold thinking together, and notice it was built for the opposite job. Your notes app, your team's wiki, the search box over a decade of documents: every one runs on a single assumption, that the task is capture and retrieval. Get the thought in. Get it back when you ask. Storage keeps the thought and drops the thread: the live relation between thoughts that carried the meaning, the reason this note mattered next to that one. Retrieval hands back the note and lets the relation fall, because it was built to store items, and a relation is the connection between them, the one thing the box has no slot for.

Vannevar Bush saw the mismatch in 1945. The mind, he wrote, works by association, leaping from one thing to the next along a web of trails it builds as it goes, while our filing systems work by rigid index, one item in one place under one name. He wanted a machine that worked by association, the way the mind does, building trails between items and following them later, a true extension of memory.4 Borges drew the nightmare version: a library holding every book that could ever be written, every page preserved and shelved, and useless for it, the one true sentence you need lost somewhere on its infinite shelves among all the false ones, the meaning drowned in the completeness of the storage.5 The tools in your pocket are smaller, and they fail the same way.

The distinction underneath all of this is small to state and decides what you build. A storage system optimizes for accuracy: is this the same as what we filed? A coherence system optimizes for something harder to measure and worth more: does this still hold together, does it still mean, now that scale and time and ordinary forgetting have worked on it? Those are different questions. They produce different machines. Almost everything on the market answers the first one well, and the problem that breaks teams and minds lives inside the second.

This runs deeper than a complaint about software, and the depth is old. An archive is an instrument of selection. It decides what counts as a record at all: it keeps what can be written down and filed, and it lets go of what has to be performed or spoken or held between people to exist.6 Whole orders of knowledge fall through that gap because they could not be inscribed, valuable the whole time. I spent years on this in doctoral work, on the way Western thinking privileges what can be seen and stored over what is carried in the body and the passing moment, and I came out of it holding a conviction I could not yet name: any system that keeps only what it can file will lose the meaning that lives in everything outside the file.

I learned the same lesson again from the opposite end, this time with no theory attached. For seven years I worked inside the coordination experiments of the Web3 world, the ones trying to fund and govern shared work with no company around it: Moloch, Gitcoin, DAOhaus, Allo. The failure I watched, over and over, lived in language itself. The incentives lined up and the information flowed; the words were where it broke. A room would agree on a word and disband, and weeks later, when the work came apart, everyone would discover they had each bound that word to a different meaning the whole time. Same word, different meanings, and the gap between them invisible until it was expensive.

These were two roads to one place. A dissertation on what archives discard and seven years watching coordination dissolve into mismatched language are a single problem approached from two sides: meaning needs holding to survive scale. Reaching it twice, by routes that share nothing, is what earns my trust. A single arrival can be self-persuasion; two independent ones are evidence.

By late 2023 the conviction had somewhere physical to live. I had two thousand markdown files: notes and transcripts and arguments I never finished, more meaning accumulated than any tool I owned could hold together. The files were fine. The trouble was that nothing could hold them together, and it sat on my own disk, in plain sight.

So I built against it. Existential reads files like those, the markdown a person already keeps on their own machine, and treats them as ground truth, leaving them yours and untouched. It reads them continuously, assesses how load-bearing each piece of thinking is inside the whole, and works to keep that whole coherent as it grows. And it refuses, by design, to reduce the thinking to a number for its worth, because the moment you measure the thing directly you begin optimizing the measurement, and the meaning rots behind it. A storage system asks whether this is the same as what you filed. This one asks whether it still coheres. The same two thousand files that had outgrown every tool I owned became the ground the answer stands on.

The problem is coherence under scale. Almost every tool we think with was built to store and retrieve what was already understood, while the work that breaks us is holding understanding together as it moves. Build for that, and you are building a different kind of thing.

Footnotes

  1. On Priestley and Lavoisier interpreting the same gas through incompatible frameworks (dephlogisticated air versus oxygen), and on incommensurability as the failure of shared meaning across paradigms: Kuhn, T. S. (1962). The Structure of Scientific Revolutions. Priestley isolated the gas in 1774; Lavoisier named oxygen and articulated the modern account. ↩

  2. The Sokal affair: in 1996 the physicist Alan Sokal published "Transgressing the Boundaries: Towards a Transformative Hermeneutics of Quantum Gravity," a deliberately nonsensical article, in the cultural-studies journal Social Text (Spring/Summer 1996), disclosing the hoax on publication. ↩

  3. Open Science Collaboration (2015). "Estimating the Reproducibility of Psychological Science." Science, 349(6251). Of roughly 100 replication attempts, fewer than half reproduced the original significant result. ↩

  4. Bush, V. (1945). "As We May Think." The Atlantic Monthly, July 1945. Bush proposed the memex, a device for "selection by association, rather than by indexing," modeled on how memory moves along associative trails. ↩

  5. Borges, J. L. (1941). "The Library of Babel" (La biblioteca de Babel). The library contains every possible book and is rendered useless by its own completeness: the true volumes exist but cannot be found among the overwhelming majority of meaningless ones. ↩

  6. On the archive as an apparatus that determines what counts as a record and privileges the inscribable over the embodied: Foucault, M. (1972 [1969]). The Archaeology of Knowledge; Derrida, J. (1996 [1995]). Archive Fever; Taylor, D. (2003). The Archive and the Repertoire: Performing Cultural Memory in the Americas. ↩