Picture everything humanity knows about playing music split into two piles. In one: every score ever written, preserved note-for-note in a vault. In the other: the living musicians who can read those scores and bring them to life. Each pile is useless without the other. Burn the scores and the musicians drift a little further from the original with every performance; lose the musicians and the vault is just paper, perfectly preserved and perfectly unreadable. Keeping hard-won knowledge alive in a society works like this. It takes two separate jobs, holding the complex original somewhere intact and keeping a supply of people trained to read it, and when a culture starts losing its grip on what it knows, it almost always loses one of the two first, the more dangerous one.

The myths note reached this problem and stopped: a scaled myth drops its decoding key, and the network either grows the institutional repair that re-supplies it or segments until it is no longer one network. Harari’s Nexus handed the book that repair under the name bureaucracy, which the book uses narrowly (see the glossary) to mean the re-supply function, not the broader machinery of real-world bureaucracy. On inspection the word is doing two jobs even inside that narrow reading, and Part IV can’t be argued cleanly until they are pulled apart.

One word, two jobs

A scaled myth is a stripped inner message. The full theology, the full law, the complex form the myth was compressed from, is no longer reaching most of the network, and its interpretive latitude has widened until a surface agreement sits on top of a dozen incompatible decodings. Carrying the dropped key back decomposes into two things the myth lost when it compressed. The first is the complex form itself, the full structure with all its qualifications and edge cases; if that is held nowhere, there is no key left to re-supply. The second is the capacity in receivers to decode at that resolution; even if the complex form sits intact in an archive, a receiver who has only held the compressed myth cannot read it. The key has to be re-installed in people, not merely re-shelved in books.

These are two operations. Hold the complex form somewhere: preservation. Install the capacity to read it in receivers: training. “Bureaucracy” was one word for a pair of distinct institutional functions, done by different kinds of institution at once, one to hold the un-compressed form alive and one to keep producing receivers who can read it. Either alone fails, and they fail differently when starved and differently when captured. The pair is older than the word: monasteries copied texts and sermons trained the laity; universities did research and ran seminars; common-law courts held the precedent and clerkships trained the bench that would hold it next. Bundled in the successful cases, conceptually distinct in every case.

The preservation function

Preservation answers where does the complex form live? Its vehicles vary (monasteries, libraries, peer-reviewed journals, lifetime academics, working clergy, a body of legal precedent, version-controlled source code as a system’s spec) but the function is one: a place where the un-compressed form keeps existing, in the heads or on the shelves of people who can afford to hold it. What makes it expensive is what makes the complex form complex. A receiver who holds the full outer message for a body of work has paid years of attention to it, cannot be mass-produced, and cannot be employed cheaply, because the time they spent acquiring the key is sunk in them. So preservation institutions, when they work, are staffed by people whose primary product is not output the network can immediately use but the continued existence of a form the network cannot itself afford to hold. That is awkward to fund, and most attempts to economize amount to insisting the preservers also popularize, which collapses the function into something cheaper to carry.

Preservation as a function is distinct from the forms it takes. The curation layer is a form (open submission, layered review, reputation-weighted moderation) that can host preservation but doesn’t constitute it: a lone academic with no apparatus performs preservation, while a well-run curation layer fails at it if it ranks the engagement-friendly variant as readily as the complex one. And preservation has three failure modes. It can be starved: defunded, its holders aging out, its archives bit-rotting. It can be captured: its holders selected for loyalty to a reading rather than fidelity to the form. And it can be displaced: the function nominally moved to a vehicle (a Wikipedia article, an LLM weight) that holds the compressed myth in nicer vocabulary while everyone proceeds as if the complex form is safe. Displacement is the failure that looks least like failure, which is why it is easiest to do and slowest to notice.

The training function

Training answers how does the key get re-installed in receivers? Its vehicles also vary (apprenticeship, schooling, graduate programs, deliberate cross-disciplinary exposure, sabbaticals) but the function is one: a place where receivers acquire a decoding capacity they would not otherwise get. The naive specification is to teach field-general capacities (statistical literacy, critical reading), and the bridge-node note worked through why that alone doesn’t produce the receivers integration needs. Field-general tools are necessary but not sufficient, because they don’t produce the chunked perception that lets a receiver hold a whole complex argument at once. Flash a chess master a real game position for five seconds and they rebuild the whole board, because they see a few meaningful formations where a beginner sees thirty-two scattered pieces (the classic result is Chase and Simon’s, 1973). That perceiving-in-large-units is most of what expertise is, and generalist tools without it are like knowing every rule of chess and still seeing only scattered pieces.

So training’s refined target is receivers who will hold the un-compressed form, not receivers who will read the compressed myth a little smarter: the rare deep specialists who have also trained metacognitive flexibility (versatile experts), and the much larger class trained at least to register the frame message, to recognize that what they’re reading has an outer message they don’t hold rather than take the bare claim at face value. A receiver who can read the frame reaches for the institution that holds the key instead of decoding the myth with whatever local key falls to hand. (Whether producing those two populations is one training function or two is a real question I flag at the end.) Training’s failure modes shadow preservation’s. Starved: credentialism replaces depth, contact time shrinks, apprenticeship withers. Captured: the curriculum installs a key chosen by whoever owns the institution. Displaced: teaching-to-the-test, or an LLM tutor tuned to agree rather than to grow the receiver. Pope Leo XIV’s Magnifica Humanitas (2026) lands on this prescriptive line, calling for “an educational alliance for the digital age” and naming “the central role of schools” as a counterforce, which grounds the training-side argument in a cross-tradition conversation rather than leaving it the book’s alone.

Why each alone fails

Run the pair as one system and the two failures hand each other their consequences. Preservation without training: the complex form is alive somewhere but can’t be reached. The monastery copies the manuscripts and no one reads Latin; the journal publishes the rigorous result and no one can parse a regression; the network runs on the compressed myth as if the form weren’t there, and defenders point at the archive and say the truth is right there. The pathology is inaccessibility, not loss, and from the network’s point of view an inaccessible form is indistinguishable from an absent one. Training without preservation: receivers are competent to decode but there is nothing to decode. Pedagogy disconnected from a living research tradition becomes pedagogy of the myth itself, with vocabulary that feels like depth but is no longer anchored in a form anyone is developing, and the receiver, because the training felt real, mistakes their competence for the kind that would notice the form was lost. The first produces apathy; the second produces confident error; the integration this book wants needs neither.

The two functions form a pump. Preservation holds pressure, keeping the un-compressed form alive so there’s something to draw from; training releases it, producing receivers who can pull from the preserved form rather than the myth. Pressure without release accumulates: the form sits unread and the institution ages into irrelevance. Release without pressure pushes thin air: receivers are trained to draw from an upstream with no water in it. The prescription, when it lands, is not to pick a side but to design paired institutions that hold pressure and release it at once, sustained by the same political settlement.

The asymmetry, and its test

The myths note left the dilution loop with three outcomes for the repairing institution: effective and credibly neutral (the loop compounds safely), too weak (fragmentation), or effective but captured (the stable bad case, scale that looks like success while distributing a tuned key). Applied to the pair, the first two are symmetric across the halves, but the captured case is not. The two halves of the pair are asymmetric under capture: captured training does more damage than captured preservation, and recovers more slowly. (The unified treatment lives in the capture taxonomy, where this generalizes: receiver-training is a consumer-key substrate and preservation a surface one, and consumer-key captures are systematically harder to recover from.) You can replace a corrupted book; you cannot replace a corrupted generation, not quickly and not from outside. A captured preservation institution holds a tuned version of the form, but the form is in principle re-discoverable: other archives exist, the evidence is still out there, outsiders can re-derive it. A captured training institution tunes the receivers themselves, and the perceptual mode they were installed with is not a belief they can revise on better evidence, it is the machinery they now read with. Two generations trained on a captured curriculum are a re-tuned readership that later un-tuned preservation has to do all its work against; the preservation can be perfect and still not land.

The claim needs testing against real cases, and the cases both support and complicate it. Soviet biology under Lysenko is the cautionary one: a captured training apparatus taught a generation of biologists a false genetics, and Soviet life science took decades to recover even after the official line changed, because the readership had been built to decode the captured key, which is the asymmetry doing what it predicts. But Lysenkoism also captured preservation, suppressing and destroying the genuine genetics literature, so it is not a clean test of training-capture alone, and it cuts against the tidy version of the thesis: preservation there did not quietly recover on its own. The Reformation runs the other way, a captured preservation half (the official canon and its authorized readings) against a partially uncaptured training half (private scholarship, vernacular literacy) where the preserved form was renegotiated faster than a pure recovery-difficulty model expects. So the asymmetry is real in the direction it points, captured training is stickier because each cohort trains the next, but it is a strong tendency rather than a law, and the clean cases are rarer than the argument would like.

There is also a decay-rate asymmetry underneath. Preservation degrades slowly: a library can sit on shelves for decades on institutional inertia, the loss invisible until someone tries to use the form. Training degrades within a generation: every cohort has to be carried fresh, and the moment the producers of versatile experts stop, the supply runs out within a working lifetime. So training is the more fragile half on the time axis and the less recoverable half on the capture axis, which makes it the more urgent half to defend, even though preservation is what the institution rhetorically advertises. Defunding patterns can be read against this: which half is being cut tells you which kind of damage the funder is choosing.

What this means for the modern environment

Once the functions are pulled apart, the diagnosis gets specific: the modern collapse is not uniform across the pair. Preservation institutions have weakened but mostly held, most cleanly at the endowed, elite core, with broader preservation collapsing where that funding is absent; training institutions have been more thoroughly weakened and more thoroughly captured. Universities are the worked case in both directions. The research side has held, with adjustments: the journals still come out, the form is still being extended. The teaching side has visibly degraded, with contact hours per student falling, contingent labor replacing the apprenticeship-style mentoring that produced versatile experts, and credentialism displacing training-for-depth. The pattern is not “the institution rotted” but “the training half rotted while the preservation half persists, and from outside the institution still looks alive because the preservation half is what advertises.” Journalism is sharper: investigative reporting (some reporters still hold the complex form of a beat) survives in expensive pockets, while the newsrooms that produced reporters with that depth have shrunk hard, US newsroom employment down roughly a quarter since 2008 and newspaper newsrooms by more than half, on Pew’s counts.

This is where the medium-shapes-want loop turns up. Training is the explicit version of what the medium does implicitly: it installs a decoding key and builds the want that pulls on the complex form rather than the myth. When training weakens, a generation’s receiver-formation devolves wholesale to the medium’s want-loop, which has by then been tuned to engagement rather than to decoding the form. The collapse of training is what hands the receiver entirely to the medium, and that is the modern environment in one sentence. It also sharpens Chapter 10’s question. A network whose preservation is intact and whose training is captured is one whose complex form keeps existing but whose receivers cannot reach it without the captured intermediary, and that intermediary becomes a chokepoint with monopoly rents. That is a structure with a clear beneficiary, not a side effect of underfunding, and it is what Chapter 10’s political economy has to take as input and Chapter 12’s infrastructure has to repair.

Where I land

“Bureaucracy” is one word for two functions that real institutions sometimes bundle and the modern environment has been unbundling. Preservation keeps the un-compressed form alive; training re-installs the capacity to decode it. Either alone fails, preservation without training into an inaccessible archive, training without preservation into competence pointed at the wrong target, and they fail asymmetrically under capture: captured training does more damage and recovers more slowly, because trained receivers cannot easily be un-trained. The asymmetry says where to defend first, and the diagnosis of the modern collapse is not uniform rot but a training half hollowed under the cover of a preservation half that still mostly works.

Where I’m still uncertain

  • Whether the two are separable in practice, or always co-bundled in any institution that works. Maybe a healthy carrier always does both because each keeps the other honest: preservation checked by the trained receivers who use it, training anchored by the form being preserved. If so, the unbundling I describe is not a feature of the modern environment but its disease, and the prescription has to insist on re-bundling rather than resourcing the halves independently.
  • Whether bridge-node production is a third function rather than a corner of training. Producing the many receivers competent to recognize an outer message they don’t hold is one thing; producing the rare versatile experts who carry the form across networks is another. If they need different infrastructures, the split has three functions, and bridge-node production has been hiding inside “training” the way the two functions hid inside “bureaucracy.”
  • The decay-rate carving is a rule of thumb, not a law. Preservation can decay fast (format obsolescence, lost languages) and training can outlast a generation (apprenticeship traditions that train trainers for centuries). “Preservation slow, training fast” is approximately right and the right strategic emphasis, but not structural.
  • The “bureaucracy” stipulation is provisional. Using the word narrowly works for the argument but doesn’t settle whether the narrow function is the core of bureaucracy or the slice needed here. The cleaner alternative, calling it institutional carriers throughout and reserving “bureaucracy” for Harari’s broad sense, is worth reconsidering once Part IV is fully in hand.

← Interlude: Myths at Scale · Chapter 8b: Intersubjective Truth →