Culture Has Read/Write Permissions
subtitle
Sampling, undercurrents, and why predicting the zeitgeist might mostly be a signal-processing problem.
In June 2023 I wrote that there was “an intuitive kind of synthesis in most sample based arts that the brain is pretty good at” and that computation was beginning to augment it. Two years later I apparently had not stopped thinking about the same thing, because I tried to explain sampling by giving the Mona Lisa filesystem permissions.
sampling in music is cool because u have a "signal" (coherent idea made of sound aka the sample itself) that is then augmented and u can repeat this ad infinitum. imagine doing this with an oil canvas. take the mona lisa and paint on top of it. certain parts keep read/write perms
The idea was that a sample is already a coherent signal. You do not begin from an empty canvas and invent every constituent part; you inherit an object that already contains structure, history, associations, rhythm, timbre, whatever, then decide which properties remain fixed and which properties you are allowed to modify. Certain parts keep read/write permissions.
That sounds like a description of one technique in electronic music. I increasingly think it is closer to a description of culture itself.
I. The sample is not the atom
Sampling is unusually revealing because it makes derivation literal. You can hear the old record inside the new one. The ancestry has not been politely disguised as inspiration; somebody took a section of an existing signal, copied it, and began operating on it. Pitch it, stretch it, filter it, reverse it, chop it into pieces, put drums underneath it, use three hundred milliseconds of it, etc. The resulting thing can be radically different without ever passing through a state of total originality.
Most other culture behaves this way too; it is simply worse at admitting it. A painter inherits perspective, composition, pigments, symbols, genres and centuries of prior paintings. An architect does not independently derive the arch. A fashion designer inherits the human body, textile technology, silhouettes, military uniforms, workwear, subcultures and whatever seventeen-year-olds in some city decided looked cool six months ago. Language is the funniest case because it is almost entirely sampled material and nobody finds it suspicious when a writer uses all the same words as everybody else.
Culture is recombinant by default. Digital sampling just exposed the operation.
There is a tempting but slightly wrong conclusion here, which is that the fundamental unit of culture must therefore be the sample, meme, motif, sign or some other little transmissible object. I think the more interesting unit is the transformation.
The same sample given to ten producers does not produce ten copies. It produces ten demonstrations of what each person considered invariant and what each person considered mutable. One preserves the vocal and destroys the harmony; another keeps two drum hits; another notices some incidental room noise nobody else cared about and turns it into the entire track. The source matters, but the read/write mask matters more. That choice is authorship.
II. Compute makes more things behave like samples
This matters now because computation is making more media behave like sampled audio. Photographs acquired layers, masks and selections. Video acquired timelines and compositing. Code was already absurdly recombinant. Generative systems take the idea further because the editable unit no longer has to correspond neatly to a region of pixels or seconds of waveform. Increasingly you can preserve pose while altering subject, preserve composition while altering material, preserve cadence while changing language, preserve geometry while regenerating texture, preserve the thing you mean while rewriting some of the means by which it is expressed.
That does not mean a generative model is literally a sampler with extra steps. The technical mechanisms are different and that distinction matters. The continuity I care about is at the level of the human interface: more of an artifact can be treated as selectively inheritable structure.
The Mona Lisa gets permissions.
This is probably why I was optimistic about computation augmenting intuitive synthesis in 2023. Humans are extremely good at the vague operation of “I want this part of that, except like this.” We are substantially worse at specifying the complete intermediate representation required to manufacture it. Traditional tools make us translate intuition into a long sequence of operations. Better computational tools compress that distance.
The scarce thing moves away from manual construction and toward selection, judgment, constraint and taste. That may make authorship more obvious rather than less. When production becomes cheap, “why did you choose this?” gets harder to hide behind technical difficulty.
III. The zeitgeist is not a leaderboard
Two days before the 2023 sampling thought, I wrote something else:
imagine cultural waves plotted across time — now Pick a point and tell me where it’s going
And a month before that:
Everyone deep down wants to be a part of the undercurrents that become larger waves in the zeitgeist
I did not have the Cultural Volatility Index name yet. Apparently I had most of the disease.
The normal way to quantify culture is to wait until something becomes easily measurable and then count it: streams, sales, searches, mentions, followers, views. This is useful, but it mostly tells you where the water already is. A trend is a crest seen from shore.
If the actual question is “where is this going?”, level is less interesting than motion. You care about velocity, acceleration, direction, mutation, transmission between communities and whether apparently separate phenomena are actually coupled. This is where the wave metaphor stops being decorative and starts suggesting instrumentation.
Not because human culture secretly obeys the Navier–Stokes equations, unfortunately. The useful part is the questions the metaphor forces you to ask. Where is energy entering the system? Which disturbances decay immediately and which propagate? Which communities are in phase? Which are leading others? When does a signal cross from one medium into another — music into clothing, clothing into language, language into advertising, advertising back into music? What appears to be noise until several unrelated sensors begin moving together?
A Google Trends chart tells you that something became popular. A cultural instrument should ideally tell you that several weak signals have begun to cohere.
IV. The people in the undercurrent
The line about everybody wanting to be “part of the undercurrents” now seems more important to me than simple hipster behavior. There is obviously status in knowing something before everyone else. People like being early. But there is another privilege available before a cultural object hardens into mass culture: you can still affect it.
At small scale the relationship between audience and creator is porous. Scenes have conversations. People steal from each other rapidly. A producer hears a track on Tuesday, does something slightly different on Thursday, and somebody else responds on Saturday. Slang mutates among ten people before ten million people repeat the finalized version without knowing why it means what it means.
The undercurrent is attractive because participation is causal there. Once the thing becomes a giant wave, most people encounter it as consumers. Beforehand, somebody got to push.
This also suggests a better definition of the “live players” I later wanted the CVI to find. A live player is not merely somebody popular; popularity can lag the property we care about by years. A live player is a node whose output meaningfully changes what adjacent nodes do. Some people receive cultural signal and reproduce it. Some distort it in useful ways. Some combine two previously separate currents and create an interference pattern that becomes its own thing. Some seem to have an unusually high gain between what they notice and what the surrounding culture subsequently notices.
Those are different phenomena, and follower count is a fairly shit proxy for all of them.
V. A Cultural Volatility Index would measure derivatives
By November 2024 I had gotten closer to the shape of it:
A volatility surface can be thought of as the surface of the ocean and underworld money and its subsidiaries understand the wave mechanics and propagation of culture etc etc you have heard it before but I’m liking the ocean vibe a little bit and it’s not even really that we’re focusing on the ocean it’s really just that oceans have waves and currents and so do markets and culture and like people as a mass as a group do yk I see a volatility surface undulating cresting
There is a reason financial volatility kept providing the analogy. Price alone does not describe a market. Two assets can arrive at the same price through completely different trajectories and therefore represent completely different states of the world. A static snapshot throws away motion.
Culture has the same problem. Imagine two artists with identical monthly listeners. One has been at that level for five years. The other had eight hundred listeners six months ago, is suddenly appearing in three geographically disconnected scenes, has had a visual motif copied by two fashion accounts, and has inspired a word people have begun using without knowing where it came from. Those are not the same cultural object. The second one has derivatives.
A serious cultural forecasting system would care less about “what is hottest?” than about propagation velocity, cross-domain transfer, mutation rate, geographic diffusion, the appearance of independent derivative works, the ratio of creators to spectators around a signal, and whether the people adopting it themselves have histories of being early to other signals. None of those individually predicts the future. Together they begin to describe a field rather than a leaderboard.
Then the old prompt becomes computationally meaningful: pick a point and tell me where it’s going.
VI. Compute makes culture more writable and more measurable
Two computational developments are colliding here. The first is that cultural production is becoming more sample-like: more properties of an artifact can be separated, preserved, transformed and recombined. Things once bound together by the physical medium become manipulable dimensions. The second is that cultural consumption and creation leave ridiculous quantities of timestamped exhaust.
Those developments feed each other. Culture becomes increasingly recombinant at exactly the moment when its recombination becomes increasingly observable.
This does not automatically produce insight. Usually it produces dashboards. The interesting possibility is to reconstruct lineage: not only that two things resemble each other, but approximately how an idea moved, where it mutated, which people repeatedly appear near the beginning of such movements and which transformations unlocked subsequent ones.
At that point cultural forecasting starts looking less like trend prediction and more like state estimation. You would never know the complete state; the system is too large, partly hidden, reflexive and full of humans who will deliberately change their behavior because they noticed you noticing them. Good. Financial markets have survived this indignity for some time.
The objective is not prophecy. It is improving your estimate of the present enough that the near future becomes less mysterious.
VII. Taste becomes more important as machinery improves
If all of this is right, computational augmentation can also do the exact opposite of what I find exciting about sampling. Sampling is powerful because a person notices something specific. The source may be common, but the read/write mask is idiosyncratic; taste determines what survives.
A system trained toward the statistical center of enormous quantities of existing culture can make recombination cheap while making selection boring. We could acquire infinite expressive bandwidth and use it to converge on the same fucking Pinterest board.
That is not an argument against the machinery. It is an argument that taste becomes more important as machinery improves. When everyone can mutate everything, cultural value moves toward deciding which mutation deserves to exist.
The live player is therefore not necessarily the person with the most technical facility. It is the person whose transformations open branches other people did not know were available. That is what a great sample does: after you hear it, the source can become permanently altered in your head. The new work reaches backward and changes the affordances of the old one.
VIII. The undercurrent is Kairos
This is also, annoyingly, connected to the essay about Kairos and Chronos. A cultural wave looks obvious after it crests. Retrospectives make movements seem inevitable because they can discard all the neighboring currents that went nowhere. The successful lineage becomes history and the dead branches disappear.
Beforehand, the system looks noisy. That is the actual object of interest.
The undercurrent is culture while its future remains negotiable. People are still sampling, mutating, choosing, combining and accidentally deciding what the eventual thing will mean. The crest is what remains once enough of those decisions have accumulated that everybody can finally name the movement. By then you no longer need a Cultural Volatility Index to tell you it exists. You need Spotify.
The thing I apparently wanted in 2023 was an instrument for the interval before that: take the incoherent field of tiny cultural movements, find the places where signal is beginning to emerge from noise, identify the people who keep showing up there, and estimate where the disturbance propagates next.
Not because culture is reducible to a waveform, but because culture is something people continuously read, selectively overwrite and pass forward. Every once in a while, if you catch the right signal early enough, you get write access.