Daniel sent us this one — he's been thinking about libraries. Not just the building with books, but the whole enterprise of taxonomy, classification, the quiet human work of deciding where everything goes. He's asking what the first library actually was, how librarians have organized human knowledge across history, whether that organization has gotten better or just different, and — the big one — what libraries even become when physical books are no longer the center of gravity. Do libraries imply paper, or can the idea survive a world where almost nothing gets printed?
Oh, this is a fantastic question. And the answer to "what was the first library" is going to depend entirely on what you're willing to count as a library. If you mean a deliberately organized collection of written documents with some kind of system — a catalog, a classification scheme — then the oldest one we have actual archaeological evidence for is at Ebla, in modern Syria. Around twenty-five hundred BCE. They found over fifteen thousand clay tablets arranged on wooden shelves, and here's the part that gets me — the tablets had these little clay tags on the edges describing what was on them. They were spine labels. Four and a half thousand years ago, someone was sitting there making spine labels.
The first librarian was also the first person to get annoyed when someone put a tablet back in the wrong place.
And Ebla had a real classification system. Administrative records in one section, literary texts in another, legal documents, lexicons — they even had bilingual Sumerian-Eblaite word lists. This wasn't just a pile of receipts. Someone had thought hard about how to structure a collection so you could find things.
The instinct to classify isn't some modern invention. It's as old as writing itself. The moment you have more than a handful of documents, you need a system. But here's what I want to know — how much of that early organization was practical, "where's the grain inventory," and how much was genuinely about preserving knowledge for its own sake?
That's the tension that runs through the whole history. At Ebla it was mostly administrative. But by the time you get to Ashurbanipal's library at Nineveh, around the seventh century BCE, you're seeing something different. Ashurbanipal was the last great king of the Neo-Assyrian Empire, and he was obsessed with collecting. He sent scribes out across Mesopotamia specifically to copy texts — omens, medical treatises, epic poetry, astronomical tables. He wanted everything. The library held something like thirty thousand clay tablets, and he boasted in inscriptions that he could read languages from before the flood. He was the first person we know of who treated a library as a comprehensive repository of human knowledge rather than just a working archive.
"Before the flood." So he's claiming to read antediluvian texts. Which is either incredibly impressive or the ancient equivalent of someone saying they've read Finnegans Wake in the original.
The Epic of Gilgamesh tablets were found at Nineveh. That's not nothing. And the library had a catalog — the tablets were organized by subject, with colophons identifying the text, the scribe, and sometimes the source tablet they'd been copied from. It was a serious intellectual enterprise. But here's where the story takes a turn that I think is instructive for the whole history. Nineveh was sacked in six twelve BCE. The library burned. And paradoxically, the fire is what preserved the tablets — they were baked hard rather than crumbling back to clay. Destruction as preservation. The knowledge survived precisely because of the catastrophe that was supposed to destroy it.
That's almost too perfect as a metaphor. But it raises a genuine question about the physicality of libraries. Clay tablets survive fire. Papyrus rots in humidity. Vellum lasts centuries but goats are a limited resource. The medium shapes what survives, which shapes what we think ancient people knew. We have tons of Mesopotamian administrative records because clay is durable. We have almost nothing from the Library of Alexandria. And yet Alexandria is the one everyone's heard of.
Right, and Alexandria deserves its reputation but also needs some demystifying. Founded around three hundred BCE under Ptolemy the First or Second — scholars argue about exactly when — it was part of the Musaeum, the temple to the Muses, which was more like a research institute than what we picture as a library. They had lecture halls, gardens, a zoo, living quarters for scholars. The library itself may have held somewhere between two hundred thousand and seven hundred thousand scrolls at its peak. And they weren't just collecting — they were aggressively acquiring. Ptolemy the Third supposedly required every ship that docked at Alexandria to surrender any books on board. The library would copy them, keep the originals, and return the copies. They literally invented the concept of "we're keeping this.
The original "you'll get it back, we promise.
They didn't just hoard. They invented bibliography as a discipline. Callimachus, a scholar-poet working there in the third century BCE, produced the Pinakes — a hundred and twenty volume catalog of the library's holdings. It organized works by genre — epic, tragedy, comedy, philosophy, medicine, law — and within each genre, authors were listed alphabetically, with biographical notes, a list of their works, and the opening line of each. This was the first attempt to systematically map all of Greek literature. It was a metadata project on a scale that wouldn't be attempted again for centuries.
A hundred and twenty volumes just to describe what they had. That's the thing about libraries — the catalog becomes its own document, its own intellectual artifact. And Callimachus's Pinakes is a perfect example of taxonomy as creative work. He's not just listing books. He's deciding what counts as epic versus lyric, who belongs in which category, what the boundaries of "philosophy" even are. Those decisions shape how later generations understand the literature. The classifier has enormous power.
And you can trace that power through the whole history of library classification. The medieval period is especially interesting because monasteries became the custodians of written knowledge in Europe after the Roman Empire collapsed. Monastic libraries were often tiny — a few hundred manuscripts was considered a major collection — but the organizational principle shifted entirely. Instead of Callimachus's genre-based system, you got organization by religious importance. Scripture first, then commentaries on scripture, then the Church Fathers, then secular works grudgingly at the end. The classification scheme encoded a theology.
The shelf order was an argument about what mattered. Which means every classification system is an argument about what matters. Dewey Decimal, Library of Congress, the arrangement of a bookstore — they're all value judgments dressed up as neutral systems.
Dewey is the most revealing example of this, because it pretends so hard to be objective. Melvil Dewey published his Decimal Classification in eighteen seventy-six. Ten main classes, each divided into ten divisions, each division into ten sections. It's elegant, it's scalable, it's still used in more than a hundred thirty-five countries. But look at the ten main classes. The two hundreds are religion, and within religion, the two-twenties are specifically the Bible. Christianity gets the entire two-twenties and two-thirties. All other religions share the two-nineties. Islam, Hinduism, Buddhism, Judaism — they're all crammed into a single subdivision. The system was built by a nineteenth-century white American Protestant, and the allocation of space reflects exactly what you'd expect.
I remember noticing this in elementary school — the entire section on non-Christian religions was about six inches wide. As a kid you assume that's because those religions are smaller or less important. You don't realize the shelf itself is lying to you.
Dewey was also a pretty unsavory character personally — serial sexual harasser, anti-Semite, racist, forced out of the American Library Association he helped found. Which doesn't invalidate the classification system, but it does remind you that systems carry the fingerprints of their creators.
The fingerprints and the blind spots. So what's the alternative? What did people do before Dewey, or instead of Dewey?
The British Museum library — now the British Library — used a system based on physical placement. Books were arranged by size and acquisition order. A book's call number was literally its shelf location. If you moved the collection, every catalog entry became wrong. It was terrible for browsing but it worked if you knew exactly what you wanted and looked it up in the catalog. The French developed a system organized by broad subject areas with fixed shelf locations. Thomas Jefferson, who sold his personal library to Congress after the British burned the original Library of Congress in eighteen fourteen, used a classification scheme based on Francis Bacon's division of knowledge into Memory, Reason, and Imagination. History went under Memory, philosophy under Reason, fine arts under Imagination. It's actually a beautiful framework — it classifies based on how the human mind works rather than just what the book is about.
Bacon's tripartite division is one of those ideas that feels almost too elegant to be useful, and yet Jefferson made it work for a library of over six thousand volumes. Which brings me to something I've been wondering — when did librarianship become a profession rather than just something scholars did on the side?
The late nineteenth century is when it professionalizes, and Dewey is central to that too — he founded the first library school at Columbia in eighteen eighty-seven. But the figure I find more interesting is Charles Ammi Cutter, who developed the Cutter Expansive Classification in the eighteen eighties and nineties. Cutter thought about the user experience of the catalog in a way that was way ahead of his time. He articulated what he called the "objects" of the catalog — it should help you find a book if you know the author, or the title, or the subject. It should show you what the library has by a given author, on a given subject, in a given genre. And it should help you choose between editions or translations. That's basically the functional requirements for bibliographic records that librarians still use today, articulated in eighteen seventy-six.
He's thinking about the catalog as an interface between a person with a question and a collection of possible answers. That's not just librarianship — that's information architecture. It's what every search engine designer does now.
And Cutter's system was the basis for the Library of Congress Classification, which is what most academic libraries use. LC is less elegant than Dewey but more flexible — it uses letters and numbers, so you can insert new subjects without disrupting the whole scheme. When computer science emerged as a field, LC could slot it into the Q's — science — without having to renumber everything. Dewey would have had to wedge it in somewhere between the zeros and the ones, which is either poetic or painful depending on your tolerance for puns.
I'll allow it. But this brings us to the twentieth century, and the explosion of information that Dewey and Cutter couldn't have anticipated. The number of books published annually went from a few thousand in the eighteen hundreds to millions by the late twentieth century. The classification systems held up surprisingly well, but the real transformation was the catalog itself — from cards to computers.
The card catalog is one of those technologies that was so successful it became invisible. Standardized index cards — seven point five by twelve point five centimeters — with a hole at the bottom for a rod that kept them from being removed. You could file by author, title, and subject, creating multiple access points to the same book. It was infinitely expandable — you just added cards. And it was collaborative — multiple libraries could share catalog cards, which the Library of Congress started distributing in nineteen oh one. It was essentially a distributed, manually-updated database.
The physicality of it mattered. You could see the card catalog — it occupied space. A big wooden cabinet with dozens of drawers. The size of the catalog told you something about the size of the collection. There was a tactile experience of browsing that no search bar replicates. You'd pull out a drawer, flip through cards, and your eye would catch something adjacent to what you were looking for. Serendipity was built into the interface.
That serendipity is one of the big arguments for physical browsing, and it's not just nostalgia. There's research showing that when people browse physical shelves, they encounter and check out books they wouldn't have found through a keyword search. The shelf arrangement creates adjacency relationships that a ranked list of search results doesn't. But the trade-off is that the physical catalog is terrible at the things computers are great at — full-text search, remote access, cross-referencing across collections, updating in real time.
We've moved from clay tags to scroll catalogs to card catalogs to OPACs — online public access catalogs. And now we're in a world where the catalog is often just a search box on a website, and the collection includes e-books, databases, journals that were never printed, datasets, code repositories. Which brings me to the second half of the prompt — what does a library even become when physical books are no longer the default format?
I think we have to separate two things. One is the library as a collection of physical objects. The other is the library as a set of functions — selection, organization, preservation, access, guidance. The functions don't go away just because the objects change format. In some ways they become more important.
Say more about that. Because the obvious counterargument is: if everything is digital and searchable, and I can find whatever I want from my phone, what do I need a library for?
The counterargument assumes that "everything is digital and searchable" is true. It's not. Enormous amounts of material have never been digitized. The Google Books project scanned something like forty million volumes, which sounds like a lot until you realize that major research libraries hold hundreds of millions of unique items. And even when things are digitized, they're often behind paywalls, or in formats that aren't easily searchable, or they're orphaned works where the copyright status is unclear. The library's role as a steward of the non-digital, the hard-to-digitize, and the legally tangled isn't going away.
There's also the problem of digital preservation, which is much harder than preserving paper. A book on acid-free paper, stored in reasonable conditions, lasts centuries. A PDF on a server lasts until someone stops paying the hosting bill. The average lifespan of a URL is something like a hundred days. Link rot is real, and it means that digital collections require active, ongoing maintenance in a way that physical collections don't. A library can buy a book, put it on a shelf, and ignore it for fifty years. A digital collection needs constant attention.
The British Library had a major cyberattack in late twenty twenty-three that took down their digital systems for months. They lost access to their own digital catalog. The physical books were fine. The card catalog — if they'd kept it — would have been fine. But their entire digital infrastructure was vulnerable in a way that shelf upon shelf of paper simply isn't. That's not an argument against digital, but it's a reminder that "digital" doesn't mean "permanent." It often means the opposite.
The preservation function becomes more critical, not less, in a digital world. But what about the other functions? Selection, organization, access, guidance?
Selection is arguably the most important one, and it's the one that's least visible to users. A library doesn't just collect everything — it chooses. That's what distinguishes a library from the internet. The internet is everything, which means it's also every piece of misinformation, every outdated medical claim, every hallucinated fact from a language model. A library collection represents a set of decisions made by human beings with expertise about what's worth keeping and what isn't. That curation function is the library's core value proposition in an age of information abundance.
The filter is the product. The internet gives you a million results. The library gives you the twelve that matter.
Guidance — the reference function — is evolving in interesting ways. Public libraries in particular have become front-line social service providers in many communities. Librarians help people apply for jobs online, navigate government websites, access telehealth services. They're doing digital literacy instruction for seniors, homework help for kids, makerspace supervision. The library building has become one of the last indoor public spaces where you can exist without being expected to buy something.
That's a point that doesn't get made enough. The library is one of the few remaining third places — not home, not work — where loitering is the point. You can sit there all day and nobody asks you to leave or buy a latte. In a world that's increasingly hostile to the idea of just being somewhere without a transaction, that's radical.
It connects back to the earliest libraries in a way that's easy to miss. The Library of Alexandria wasn't just a collection of scrolls — it was a place where scholars lived and worked and argued with each other. The medieval monastery library was embedded in a community with shared practices and values. The nineteenth-century public library was explicitly framed as a tool for self-improvement and democratic citizenship. Libraries have always been social institutions, not just storage facilities.
If we project forward — say, fifty years — what does a library look like in a world where the amount of books printed on paper has become, as the prompt puts it, "almost vanishingly small"?
I think we'll see a bifurcation. Research libraries and special collections will become more like museums — places where you go to access rare and unique materials that don't exist anywhere else. The physical objects become more valuable precisely because they're scarce. A medieval manuscript has a kind of aura — to use Walter Benjamin's term — that no digital facsimile can reproduce. Scholars will still need to examine physical objects for material evidence that doesn't show up in scans.
The physical book as primary source rather than delivery mechanism. You're not there to read the text — you can read the text anywhere. You're there to see the watermarks, the binding, the marginalia, the stains that tell you which pages someone read most often.
And then public libraries and community libraries will evolve in the other direction — less about the collection as a physical asset, more about the services and the space. We're already seeing this. Libraries lend tools, seeds, musical instruments, Wi-Fi hotspots, museum passes. They host community events, citizenship classes, tax preparation help. The books are still there, but they're not the only thing, and in some branches they're not even the main thing.
There's a library in Helsinki — Oodi, the central library — that opened in twenty eighteen. It's a hundred and seventy thousand square feet, and the entire third floor is a book space, but the rest is recording studios, three-D printers, sewing machines, gaming rooms, meeting spaces, a cinema, a café. It's called a library because that's the institutional lineage, but it's really a public knowledge commons. The books are almost an afterthought.
Oodi gets about three million visits a year. For a city of six hundred fifty thousand people. That's not a dying institution. But it also raises a question about taxonomy that I think is underexplored. If a library lends sewing machines and three-D printers, how do you catalog those? The traditional classification systems are built around texts. What's the Dewey number for a drill press?
The Dewey number for a drill press is probably six twenty-one point nine, which is "power tools" under "mechanical engineering." But you're right that the system strains when you try to classify objects rather than texts. And that's the deeper question — if libraries become general-purpose knowledge-and-tool commons, do we need a whole new information architecture? Something that doesn't assume the primary unit is a codex?
There have been attempts. The semantic web movement, linked data, RDF triples — the idea that instead of organizing books on shelves, you organize facts and relationships in a machine-readable graph. The Library of Congress has been doing work on BIBFRAME, which is an attempt to replace the MARC format — MARC being the standard for machine-readable cataloging since the nineteen sixties — with something more native to the web. MARC was designed for printing catalog cards. BIBFRAME is designed for linked data on the internet. It treats works, instances, and items as separate entities with relationships between them. It's a fundamental rethinking of what a bibliographic record is.
This is where I think the history of libraries is actually deeply relevant to the future of AI and knowledge management, which is something the prompt is hinting at without quite saying it. The problems librarians have been solving for four thousand years — classification, disambiguation, authority control, preservation, retrieval — are exactly the problems that large language models and knowledge graphs are trying to solve now. Librarians invented the solutions that we're now trying to automate.
Authority control is the perfect example. In library science, authority control means establishing a single canonical form for a name or subject — so that "Twain, Mark" and "Clemens, Samuel Langhorne" point to the same author, and you don't have half the books filed under one name and half under the other. It's disambiguation. It's entity resolution. It's exactly the problem that knowledge graphs like Wikidata and Google's Knowledge Graph are solving at enormous scale. Librarians have been doing entity resolution by hand since Callimachus.
They've been doing it with a rigor that a lot of automated systems lack. A library catalog entry is a highly structured document created by a trained professional who has thought about how this particular book relates to other books, what subjects it's actually about versus what the title claims it's about, which edition it is, whether the author's name has changed, whether the work is part of a series. A language model can generate a plausible summary, but it doesn't have the systematic commitment to accuracy that a cataloger brings.
Though I'll push back slightly — the library world has also made mistakes at scale that automated systems might help correct. The Library of Congress Subject Headings, which are used by libraries across the English-speaking world, have historically been terrible at representing marginalized communities. For decades, the subject heading for LGBTQ topics was "Sexual deviation." It wasn't changed until nineteen seventy-two. The heading for Indigenous peoples was "Indians of North America" with all the inaccuracies that implies. These aren't neutral technical decisions — they're classification choices that encode bias, and they take decades to fix because the systems are so entrenched.
The future might be a hybrid — human catalogers making the high-stakes intellectual decisions about what a work is and how it relates to other works, while automated systems handle the scale problem of making all of that discoverable across collections. The librarian's judgment becomes more valuable, not less, because it's the thing the algorithm can't replicate.
I think that's right. And there's another dimension here that we haven't touched on — the library as a bulwark against censorship and the rewriting of history. When a book is only available digitally, it can be altered or removed without anyone noticing. Amazon has deleted books from Kindles remotely — it infamously removed copies of nineteen eighty-four, of all things, back in two thousand nine because of a rights issue. A physical book on a library shelf can't be remotely deleted. The library's preservation function has a political dimension.
The memory of the culture lives in libraries. And memory is fragile. We talked about the Library of Alexandria burning — but most of the loss of ancient literature wasn't a single catastrophic fire. It was slow neglect. Scrolls that weren't recopied. Texts that fell out of the curriculum. Works that survived in a single manuscript that someone used to wrap fish. Libraries are the defense against that entropy, and they're only a defense if they're maintained across generations.
Which is why the question of what libraries become isn't just a practical question about formats and budgets. It's a question about what a society is willing to preserve and why. The Ebla tablets survived because they were in a palace that burned. The monastic libraries survived because monks kept copying. The great research libraries of the nineteenth century were built by industrialists who decided that amassing knowledge was a form of legacy. Each era finds its own reason to maintain libraries. The challenge now is articulating that reason in an age that thinks everything is already on its phone.
I think the reason has to be: everything is not already on your phone. What's on your phone is a carefully selected, algorithmically mediated, commercially driven slice of human knowledge. The library is the whole messy collection, including the parts that nobody's monetized, the parts that are unfashionable, the parts that contradict what everyone currently believes. The library is the backup copy of civilization.
That's a good place to land. But I do want to mention one practical thing — the library of the future is also being shaped right now by the open access movement and the fight over who controls scholarly publishing. Academic libraries spend enormous fractions of their budgets on journal subscriptions from a handful of publishers — Elsevier, Springer, Wiley — and those subscriptions keep getting more expensive. Libraries are pushing back by supporting open access journals, preprint servers, institutional repositories. The dream is that the research that scholars produce with public funding should be freely available to the public. Libraries are the natural home for that dream.
The future library might be less about buying books and more about hosting and curating knowledge that's produced and shared openly. The librarian shifts from purchaser to publisher, from cataloger to community organizer, from guardian of the stacks to steward of the commons. Same fundamental mission, radically different tools.
The same fundamental tension — how do you organize it all so that someone can find what they need without being overwhelmed by what they don't? That's the four-thousand-year-old problem that isn't going anywhere.
I think we've earned our fun fact for the day. And now: Hilbert's daily fun fact.
Hilbert: In the eighteen forties, British naturalist Sir Robert Schomburgk was exploring Guyana when he documented something peculiar: a species of coral snake, Micrurus lemniscatus, whose venom causes a delayed neurotoxic effect that can take up to eighteen hours to manifest. The truly strange part — Schomburgk's notes record that local guides refused to handle the snake not because of its bite, but because they believed the snake had a symbiotic relationship with a specific nocturnal bird that would hunt down anyone who killed one. The bird was never identified by Western science, but the belief persisted in local oral tradition for at least a century.
The snake had an avian bounty hunter on retainer. Nature is terrifying.
That's either a remarkable piece of unstudied interspecies mutualism or the best story anyone ever made up to keep colonists from messing with local wildlife. Either way, I respect it.
Thanks to Hilbert Flumingtop for producing, as always. The question we'll leave you with: next time you're in a library — physical or digital — pay attention to how things are organized. Someone made a decision about where everything goes. That decision shapes what you find and what you don't. That person is doing work that goes back to the clay tablets of Ebla, and it's not going away just because the tablets are now servers. This has been My Weird Prompts. Find us at myweirdprompts dot com.
