A History of Narrative Environments

the size of a narrative

This essay is part of the Protocol Narratives series

When one person tells a story, you have an isolated act of narration. When a few people tell a story in a coordinated way, and also pass it on relatively intact to future generations, you have a rudimentary narrative tradition. When you have a million people trying to tell a story at once, you either get an unnarratable meltdown, or invent narrative protocols capable of containing the complexity and scale. In 2023, humanity is at this fork in the road.

The history of narrative environments begins with just two layers, put together in a relatively crude way, via a relationship of organic, largely ungoverned emergence. It has now evolved to a stack with five layers, put together in a vastly more sophisticated way, with various narrative protocols competing for dominance.

In the process, it has almost become capable of accommodating stories told by millions of people at once, and almost mediating narrative wars on a global scale. As a prelude to talking about emerging narrative protocols, in this essay, I want to trace the evolution of narrative protocols to the present.

Pre-Industrial Narrative Environments

The evolution of the pre-Industrial narrative environment can be divided into two phases: pre-Gutenberg and post-Gutenberg. The stack remained a two-layer stack — isolated acts of narration and narrative traditions — through both phases, but the relationship between the layers shifted gradually through the entire period.

As best as I can tell, narrative traditions in antiquity were sometimes initiated by elites in a top-down fashion, but for the most part emerged from isolated acts of narration being naturally selected, bottom-up, for more complex developmental attention. Depending on the degree of institutionalization and complexity of development that followed, the traditions took the form of folklore, mythologies, religious cosmogonies, or highly stylized intellectual-literary traditions.

Due to the crippling severity of technological limitations, isolated acts of narration and bottom-up narrative traditions account for almost all storytelling up to the industrial era. In pre-modernity, I’d guess the ratio was approximately 90:10. The vast majority of stories told were random one-offs that hewed to no larger logic, and were rarely even retold once, let alone incorporated into an oral culture or written down. But the minority that did hew to larger logics, and attracted conscious oral and textual preservation efforts, had a disproportionate influence on worldly affairs.

If you go to the market and witness a funny incident, and come home and narrate it to your family, that’s probably the end of it. It’s not likely to turn into a story that is remembered in a year, let alone across generations. But when a high priest carefully crafts a narrative tradition (perhaps appropriating some convenient bits of folklore in the process) about an entire monarchial history, and has it carved in stone by slaves, it can be vastly more consequential. A 10% share of narrative environments, with the leverage of institutional memory, can dominate the 90% that is ephemeral.

Much of our understanding of narrative culture is shaped by historical memories of pre-Gutenberg narrative environments. The two extremes of the spectrum of narrative traditions, folklore sustained by a people, and epic narrative, typically sustained by a state-like apparatus, constitute something like a minimum-viable narrative environment, and remain first-class citizens of narrative environments even today. The two extremes are not in fact that different. The latter is, to some degree, merely a curated and aggrandized agglomeration of the former. And both rest on the ground created by ongoing isolated acts of narration, which rely on nothing more sophisticated than oral repetition and memory to persist long enough to be mined by, and for, prevailing narrative traditions. In the pre-industrial world, narrative traditions were to isolated acts of narration what long-term memory is to short-term transactional memory.

The early modern era between the Gutenberg revolution to the mid 19th century did not, in my opinion, change the kinds of narrative that were extant, only their form of expression, and relative proportions. The ratio shifted perhaps, to something like 85:15 (these numbers, by the way, are obviously notional, but I think they are the right order of magnitude). More importantly, the balance between bottom-up and top-down narrative traditions shifted somewhat. Printing allowed more top-down narrative traditions to emerge and take root.

The overall picture did not change much by modern standards. Largely oral traditions turned into increasingly printed textual traditions, and changed somewhat in the process. For instance, verse gave way to prose, and high-context word-of-mouth transmission gradually gave way to more impersonal and low-context transmission through printed texts. The number of narrative traditions, and the distribution of power among competing traditions, also shifted. Most famously in the evolution of narrative landscape of Christianity from a monopoly to something like a pamphlet-war-mediated oligopoly. As I said, despite the injection of a powerful new technology, the stack remained two layers high: isolated acts of narration, and narrative traditions. The upper layer simply became somewhat more competitive, and gained some mass relative to the lower layer.

At the upper level of narrative traditions, more people, with less power on average, told stories, and tended to do so in prose rather than verse. They also competed with each other more aggressively. Orality gradually gave way to textuality, but I think the significance of that tends to be overstated. Printing did eventually radically reshape the narrative environment, but it would take a few more centuries to get there.

Industrial Narrative Environments

It was only in the industrial era that the world witnessed the rise of a truly new narrative environment, with the 300-year-old promise of the Gutenberg revolution being fulfilled.

This, I think, began with improved and cheaper printing techniques that emerged between 1850 and 1900, giving rise to all the familiar print media of today: paperback novels, magazines, newspapers. This was the “Big Data” moment in the history of narrative cultures. The volume, velocity, and variety of narrative production all increased suddenly and dramatically. At the same time, the conceptual technology of storytelling evolved qualitatively beyond narrative traditions in the pre-modern sense. The evolution was driven by the nature of the new media. Each new industrial technology that came after print — cinema, radio, and television — expanded the sophistication, reach, and power of the new narrative environment.

Perhaps the biggest shift in the conceptual technology of storytelling was from a mostly zero-sum view of narrative to a positive-sum one. In the pre-industrial era, the narrative environment was something like a historical store mined from everyday life, and carried forward through generations. There was no strong distinction between the functions of storytelling and history. New stories were added to the store roughly at the rate at which history itself unfolded. Remembering and mythologizing were part of the same process.

The conceptual leap of the 19th century lay in the idea that, like machines and buildings, new stories could be conceived through an act of imagination and produced and distributed. The rate of narrative production became decoupled from the rate at which history was unfolding.

There is an analogy to be made here to the lump of labor fallacy in economics, which holds that there is a fixed amount of work to be done, and that any displacement of labor by automation is a permanent loss. This fallacy still guides a great deal of thinking in labor politics, but historically, it has never been true over the long term, though it can appear true over short periods. The amount of work to be done in the economy grows with automation, usually at an accelerating rate.

Something similar is true of the narrative environment. Many people believe in what we might call a lump-of-narrative fallacy, according to which, if one story has been told, there is one fewer story that can be told. Early in the industrial era, the source material of history was viewed as the source of scarcity. More recently, “attention” has been held up as the supposed “scarce” quantity (we’ll get to why it isn’t in a bit).

The mid-19th-century conceptual leap broke this fallacy, though many have still not noticed. The rate of narrative production, and therefore the rate of evolution in the narrative environment, was no longer limited by either the pace of history, or the supply of “attention.”

When thousands of people try to tell tens of thousands of different stories at scale, with at least a cosmetic element of innovation in each, and with attention to the affordances of technological media, you get the modern media landscape. On this landscape, I would place Hollywood-style entertainment industries, journalism, statist propaganda machines, corporate PR, and the advertising industry.

The media landscape began emerging as the third layer of the narrative environment starting around 1850.

Unlike pre-modern narrative traditions, and even Gutenberg-era pamphlet-war narrative traditions, industries in this landscape typically work with a handful of reliable and scalable patterns such as The Hero’s Journey in fiction, or “Nationhood” origin myths in politics. You do not need to be a Homer or Shakespeare to use these patterns effectively. They are industrial-strength patterns that automate away much of the need for creative genius. Relatively ordinary writers can script a summer blockbuster or hit TV show. A relatively small country like South Korea can ramp up an entire narrative-industrial complex, capable of projecting soft power globally, in a matter of decades. A mediocre staff of PR professionals can manage the narrative production needs of an entire government.

The process only gets easier, as the newly minted “masses” get more literate in the tropes and conventions of industrially produced narratives, and able to receive increasingly telegraphic modes of narration. This is one of many reasons the “scarce attention” argument is fallacious. Humans get vastly more efficient at absorbing narrative inputs, even as they get better at producing it. And as narrative environments increase in sophistication, you do not need to actually be paying direct attention to be influenced by them. For example, few people are aware of, or capable of narrating, the “grand narrative” of a typical country in any coherent way. Yet that narrative is obviously embedded everywhere in the environment, with bits and pieces poking through to surface awareness where attention must be consciously directed.

The industrial media landscape did not entirely supplant what already existed. It took over and semi-automated some narrative traditions, invented some new ones, erased many (through both deliberate efforts and inadvertently, by poaching the artisan talent they required to sustain themselves), and left some alone. I suspect it barely affected the vast mass of isolated acts of narration below the level of traditions. At most, the machinery of industrial media occasionally discovered, and retold at scale, an isolated acts of narration that had the potential.

By the time the internet emerged, the 150-year-old industrial-age narrative landscape had settled. I’d guesstimate the new pie chart at something like: 80% isolated acts of narration, 2% pre-modern narrative traditions, and perhaps 18% industrial narrative production. The lowest layer of the stack did not shrink in absolute terms. It likely even expanded. But industrial narrative production simply exploded, causing isolated acts of narration to shrink in relative terms. Pre-modern narrative traditions, by contrast, now the middle rather than top layer, did shrink sharply in absolute terms, not just relative.

The Internet Narrative Environment

Then, the internet happened.

Unlike both the Gutenberg revolution, and the late-industrial mass media revolution, it turned out to be a powerful enough technology to reshape the largest slice of the pie: isolated acts of narration.

The funny incident at the market, which previously would have been narrated once, orally, to a small family-and-friends audience, could now be cheaply written down and globally distributed in media ranging from tweets and Reddit am-I-the-asshole posts to more carefully crafted blog posts. Both accidental and performed narrations could be recorded and shared as video.

These were developments that directly affected 80% of storytelling at the most atomic level imaginable. Within just a couple of decades, smartphones have formed a network of nodes that forms a vast interposing intelligence mediating our previously direct experience of the world. This interposing intelligence produces copious amounts of narrative exhaust (my generalization of the old idea of data exhaust) that gets cheaply captured on social media, with varying degrees of conscious crafting.

But this reshaping of the 80% was not the most important effect. The most important effect was the reshaping of the two higher, and more complex layers. First, narrative traditions in the pre-modern sense went online and became utterly transformed. Traditional folklore turned into WhatsApp rumor flows. Mythologies turned into QAnon-style confabulation cults. And an entirely new kind of feedstock emerged to fuel it all: memes. In a remake of The Graduate, the word plastics would be replaced by memes.

This might seem like a bold claim, but I do believe internet memes are qualitatively and technically different from superficially similar strata of narrative fragments that might have existed in pre-modern times. They also significantly overload Dawkins’ original definition of meme, for which the original prototypes were things like the opening bars of Beethoven’s 5th, to the point of unrecognizability. Memes are not like the opening bars of Beethoven’s 5th. They’re like nothing that has come before.

Memes disrupted the prevailing three-layer stack of narrative environments: isolated acts of narration, narrative traditions, and the media landscape. By 2010 or so, a new slice had been carved out of the pie.

The Meme-Vibe Layer

Memes in the modern sense are a self-aware (via their human hosts) and reflexive (in the sense that they feed on themselves) new phenomenology that constitute something like a living mass of human-in-the-loop narrative ooze.

They draw from both above and below. From “above,” memes atomize isolated shots and fragmentary reactions from familiar larger works, like movies, TV shows, and news broadcasts, as well as tropes and references from pre-modern narrative traditions. From “below,” isolated acts of narration (such as the behavior of a Karen caught on camera) get elevated to memetic status. But there is also raw, original production of pure memes, both dank and wholesome, in the cozy underground caverns of the internet, driven by an increasingly conceptually sophisticated production capability. If the industrial era witnessed a shift from zero-sum to positive-sum narrative production, the memetic era, by working with a more atomized class of primitives, has acted as a narrative accelerant, creating an open environment of interchangeable narrative parts.

Low-level vitalist memeing is so powerfully coherent, it has formed a whole new layer in the narrative stack, interposed between isolated acts of narration, and narrative traditions proper. In my linked essay, the headline is an allusion to LLVM, the prototype of a key architectural layer in modern computers. Memes constitute a similar layer in the narrative stack.

Once memes reached a particular level of sophistication, individual memes began oozifying further into near-invisibility, producing affective backdrops for foreground acts of narration; what we’ve come to refer to as vibes.

Vibes are to narratives what development environments are to computer programs. They offer a set of default dispositions that invisibly shape the nature of programs that get written, and indeed, what programs get written in the first place. We are more likely today to let our storytelling be constrained by awareness of broader vibes being “off.” Acts of narration can no longer be truly isolated. They exist within a kind of narrative matrix. This matrix is not just conceptual (shaped by literacy in memes and vibes) but literal, in that the physical events with which stories begin are increasingly captured by infrastructures of surveillance and sousveillance. Even acts of surveillance can themselves be surveilled, fueling meta-stories (for instance, I just saw a viral video of a woman catching a man in the act of taking a photo of another woman without consent, and precipitating a showdown).

The topmost level of narrative culture, the industrial media landscape, has been less transformed, but that is primarily because not enough time has passed. Nothing like the memefication of narrative traditions is visible in industrial production. While modern movies are made in ooze-aware ways (featuring many shots and set pieces obviously designed for memetic appropriation), the internal impact has been limited. Politicians use memes to some extent, but still rely mostly on old-fashioned speeches, news-making, and photo opportunities. Corporations try to run funny twitter accounts, but mostly stick to traditional PR and marketing operations.

What we do see, however, is rapidly falling effectiveness for traditional media methods. As vital streams of narrative energy sink through the base layers of the industry, like a river going underground in a shifting geology, the media landscape increasingly looks drained of life. Its output either has a mechanical, zombie-quality to it, or is marked by retreat towards low-risk, high-residual-potential old material.

Reboots, franchise extensions, and extended universes are all symptoms of a vast ongoing disruption from below. This applies not just to entertainment media, but all media. Political narratives and corporate boilerplate exhibit the same symptoms. As aggressive and memefied bottom-up narrative traditions seize control of key narratives, “official” narratives appear increasingly lifeless and toothless.

As of approximately 2018, we might summarize the state of narrative culture as follows.

1. Once-isolated acts of narration, still constituting about 70% of all narratives, are no longer isolated, but exist in a low-level matrix comprising a context of memes, vibes, and surveillance infrastructure.

2. Memes and vibes themselves constitute a thin, but hugely consequential new feedstock layer, an oozy narrative goo representing perhaps 10% of new narrative production in fragmentary form, but having a transformative influence on all other layers.

3. Narrative traditions now emerge and persist primarily in the cozyweb, fueled by the feedstock of memes and vibes, and driven by explicit and ambitious agendas that compete with those of the industrial media landscape. I’d put these at around 10%. The ground lost in the previous century and half has been rapidly regained.

4. The industrial media landscape is being disrupted in all its manifestations: Entertainment, news, politics, and corporate communications. This is currently visible primarily through the growing weakness of the landscape, and signs of awareness and accommodation of lower layers. This layer still accounts for about 10% of the narrative environment, but is in retreat. It is something like a narrative Deathstar. Potent where it can direct its energy, but unable to be everywhere, all the time.

The new four-layer stack is much more bottom-heavy, less organized, and less predictable at a beat-by-beat level (but perhaps more predictable at the level of cycles).

Into this picture, which has had almost no time to stabilize, we are now injecting yet another pair of technologies: AI and protocols. The effect of the former is more visible, charismatic, and dramatic. The effect of the latter is largely invisible, uncharismatic, and undramatic. Together, they act as something like the First and Second Foundations in Asimov’s stories, forming an emerging scaffolding for a radically restructured narrative environment, even as the old environment unravels.

The four layer stack of today is something like a transient pidgin. Not only is it too new for any of us to have gained any significant literacy in it, it is too young and unstable to even sustain any sort of refined literacy.

The emerging environment though, will be more of a creole when it matures. It mixes the elements with much greater harmony.

The World-Building Layer

Here are the five layers of the narrative environment right now, as of 2023.

1. Matrixed acts of individual narration, driven by individual imagination, observation, or experience, including both humans and AIs as originators, in both pure form, and various mixed, centaurish forms.

2. A layer of narrative ooze, comprising memes, vibes, and patterns of repetition-with-variation driven by AI tools. The well-known Balenciaga meme is a good early archetype of this future.

3. Narrative traditions that are increasingly about relations with technology (such as communities that form around particular schools of prompting), and make use of technologies (including generation, as well as other primitive operations like summarization, elaboration, interpolation, and so on).

4. A newly coalescing layer generally referred to as world-building, which includes, besides the familiar varieties in movie franchises and gaming, things like corporate imaginaries of the sort cultivated by Elon Musk around his industrial empire, and the increasingly codified partisan interpretive rubrics for political discourses.

5. An increasingly solipsistic media landscape that is grounded in generative world-building rather than reality-referencing. Early signs are already visible in the rise of extended universes, ideological bubbles of news reporting and analysis, and perhaps most subtly and significantly, emerging schools of corporate PR.

Some might argue that world-building, the new layer in the picture, has existed for a long time, but I suspect what we’ve seen so far has merely been the foreshadowing of a possibility. It takes the power of generative AI, deployed at scale, to do the vast amount of procedural construction work required to create and sustain usable narrative worlds. Importantly, and particularly in the case of world-building in corporate and political realms, the individual worlds must interoperate, forming a kind of internet of autonomous worlds. Unlike the Marvel Cinematic Universe, they cannot remain walled gardens.

In the world of entertainment, extended universes constructed out of comic-book universes, painstakingly manufactured through near-artisan levels of human labor over decades, are something like a very small store of fossil fuel. For world-building to form a new layer in the narrative environment, a cheap, renewable source of raw material is needed. One that can do in weeks or months what Stan Lee and his team of creatives took decades to do at Marvel.

This is where AI turns into a significant enabling tool. But while necessary, AI is not enough.

By itself, operating on the incumbent pidgin stack, it seems like AI can only accelerate the oozification of the narrative environment into complete unnarratability. We should see the world turn into a frenetic maelstrom of deep fakes, procedurally generated memes, misinformation, disinformation, leaks, screenshots, and empty banalities that never stabilizes long enough to allow worlds to form and persist long enough to serve even solipsistic functions. But this is not happening; at least not as rapidly as I’d expect. Invisible forces seem to be kicking in, lending structure and moderation to the process, and creating enough stability for worlds to form and persist. The oozification seems to proceeding along more orderly lines than one would expect.

This is the result of hard-to-see element in the picture: emerging narrative protocols.

Historically, the narrative stack has emerged and acquired new layers largely through unmanaged, organic processes. But for the first time, there is enough engineering agency around to consciously design the stack. Something like urbanization is happening to the narrative environment.

I opened this essay with an observation: When you have a million people trying to tell a story at once, you either get an unnarratable meltdown, or invent narrative protocols capable of containing the complexity and scale.

It appears we’re already doing the latter. In the next installment of this series, I’ll explore how that’s happening.