There Is Nothing New Under the Sun

Myth and Mechanism

The Preacher had seen everything. Vanity of vanities — hevel, breath, vapor — the endless cycling of human striving through new materials. One generation passes, another comes, the sun rises and sets, the rivers run to the sea and the sea is never full. He was not despairing. He was observing. The form recurs.

What recurs, specifically, is this: humans have always been able to do more than they could understand. The gap between capability and comprehension is not a modern condition, not a technology condition, not even a civilizational condition. It is the permanent human condition, as old as the first hand that made fire without knowing what fire was.

The Pleistocene human could observe. Lightning struck, crops failed, animals moved in patterns that resisted explanation. The observation was real and acute. The mechanism was invisible. Into that gap — between what could be seen and what could be understood — humans built myth — not false belief, but the functional structure that enables action when explanation is unavailable. Myth let people plant, coordinate, grieve, go to war, transmit knowledge across generations without requiring the underlying mechanism to be understood. The myth was not the failure. The myth was the solution.

This is the pattern the Preacher saw. Observation outruns explanation. Myth fills the gap. Myth is functional until it isn’t. Eventually the framework expands — sometimes after catastrophic cost — and the myth is replaced by mechanism. Then the cycle turns, and a new gap opens.

The history of technology is a history of that gap opening and closing, at different speeds, with different costs.

The plow closes the loop instantly. You see the furrow. The gap between action and observable consequence is a single stride. The loom opens a slightly larger one — the mechanism is less visible, but the output remains legible: a buyer can evaluate cloth without understanding the machine. No myth required in either case.

Pharmacology opens a different kind of gap entirely — not a gap in the output, but a gap in the system being acted upon. The human body is a complexity that no framework has ever fully described.

In 1957, thalidomide was approved for morning sickness across Europe, prescribed by careful physicians working at the frontier of their field. The drug worked for the stated purpose. The children born without limbs in the years that followed were not the result of negligence or corner-cutting by the standards of the framework available to them. They were the result of a framework that could not generate the question that would have mattered — not because the concept was unavailable, but because the testing regime did not require it. The question of what this compound does to a developing embryo was askable. It simply was not being asked. The framework did not contain the obligation to ask it.

The distinction matters. Frances Kelsey of the U.S. FDA asked exactly that question in 1960, found the toxicology inadequate and the efficacy evidence weak, and blocked American approval. The framework could generate the question. It generated it in one person, in one regulatory body, in one country. Everywhere else, the gap held.

This is the distinction the Preacher’s frame requires. There is prediction failure — you know what to look for and don’t look hard enough. And there is perception failure — the framework cannot see the variable that governs the outcome, because the obligation to look for it has not yet been institutionalized. Thalidomide was perception failure at the regulatory level. The science could ask the question. The system did not require it.

The gap between cause and observable consequence was a decade. The framework eventually expanded — and the expansion was not merely corrective. The thalidomide disaster helped drive a regulatory and scientific focus on systematic teratogenicity testing, and on the differential effects of stereoisomers in pharmacology. Later work on asymmetric chiral synthesis — producing single stereoisomers reliably rather than mixed forms — was recognized with the 2001 Nobel Prize in Chemistry. The same molecule, now understood, returned to clinical use as a treatment for multiple myeloma and leprosy. The myth failed, catastrophically, at the edge. The framework expanded. The tool was eventually understood well enough to be used safely for a different purpose. The edge cost was real and irreversible. The framework gain was also real.

This is the arithmetic of rapid adoption at scale: it optimizes for the mean, and accepts the edge as cost. The mean benefit of a new technology is distributed across a population. The edge cost is concentrated in specific people who did not consent to being the edge. The calculation is almost always made, and almost always correct at the population level. What it cannot do is make the edge cost not have happened.

The rhythm across the full historical record is the same. Lead pipes ran through Roman cities for centuries. The health effects were real; the mechanism was invisible. The gap between action and consequence was measured in generations, the harm distributed too slowly to be legible as harm. Asbestos was used for four thousand years. Greek and Roman writers noted that workers who handled it became ill and died young. The gap between observation — workers are dying — and mechanism — fibrous mineral particles lodge permanently in lung tissue — persisted for two millennia. Germ theory as the explanation for infectious disease was not consolidated until Pasteur and Koch in the 1860s and 1880s, roughly two thousand years after the miasma framework had established itself as the coherent, functional, apparently validated myth. John Snow removed the handle from the Broad Street pump in 1854 and the cholera outbreak stopped. The mechanism still wasn’t accepted for another generation, because correct intervention without a replacement framework does not close a gap. You need a better framework, not just better data.

Each of these gaps eventually closed. The horizon was fixed. Human understanding was slow, but the technology being understood was not getting more complex while humans tried to understand it. Lead pipes did not compound. Asbestos did not iterate. The myth was eventually replaced by mechanism, and the replacement held.

In 2001, Ray Kurzweil observed that when science becomes software, Moore’s Law takes over. The insight seems technical. Its implications are civilizational.

Every prior legibility gap was temporary because the technology being understood sat still while understanding caught up. The horizon was fixed, and human understanding — however halting, however costly in its delays — could eventually reach it. Germ theory eventually displaced miasma. The frameworks expanded.

What Kurzweil identified is a condition under which the horizon moves. When the tool itself runs on compounding computation, the distance between what the tool can do and what the framework can explain about what the tool just did does not close — it widens. The tool outruns the epistemology structurally, by design. You are not chasing a fixed point. You are chasing a point that accelerates away from you.

The limit case is not total incomprehensibility. It is the point at which comprehension becomes irrelevant to use. When a system can be reliably used without being understood, the pressure to close the legibility gap weakens. The historical cycle’s expansion phase — in which framework eventually catches up to technology — becomes optional rather than inevitable. Whether the framework expands or not no longer determines whether the technology is adopted. It only determines whether the edge costs, when they arrive, will be visible in time to matter.

The myth-structure you build today will be built for a system that is different by the time the myth is stable. The Preacher hadn’t seen that. Not because he lacked wisdom. Because it hadn’t happened yet.

Technology acceptance has always been gated by a simpler question than the legibility gap suggests. Not: do people understand how this works? But: can the incomprehensible parts be kept off their plate?

The cell phone succeeded despite the gap between its mechanism and any user’s capacity to understand it being essentially infinite. Almost no one who adopted the cell phone in the 1990s understood the difference between cellular architecture and the landline system it was replacing. What they understood was a twelve-button keypad. The decision to map cell phone operation onto the familiar landline dial pad was a design decision that functioned as a myth: a familiar story laid over an incomprehensible reality, structured so that users never had to enter the room where the complexity lived. There was one seam. Early users noticed the absence of a dial tone before dialing — a small behavioral assumption imported from landline architecture that leaked through the interface. People felt the strangeness for a moment, then adapted. The myth held everywhere except that one place where it forgot to cover itself.

When that threshold is crossed — when adoption becomes self-sustaining, when the question shifts from whether to adopt to how to adapt, when the technology is no longer optional for the culture regardless of whether individuals choose it — the snap has occurred. Resistance does not gradually erode. It collapses. The printing press snapped across Europe within fifty years; the Muslim world’s resistance held for several centuries more, sustained by a myth-structure — the sanctity of handwritten Arabic — that the European framework did not share. The automobile snapped in a couple of decades, going from marginal to dominant in U.S. transport between roughly 1910 and 1930. Anesthesia snapped almost overnight: ether was demonstrated in 1846, and within a few years surgical practice across Europe and North America had changed. The relief was so immediate, so total, so asymmetric in its benefit over the alternative — surgery without it — that resistance evaporated before it could organize.

Against this history, consider what happened on August 22, 2022, when Stability AI released Stable Diffusion to the public. ChatGPT followed in November of the same year. The gap between emergence and what can reasonably be called civilizational entrenchment — AI embedded in professional workflows, educational institutions, creative practice, research, legal and medical consultation — closed in approximately thirty-six months.

The Collingridge dilemma, named for the technology theorist David Collingridge, holds that early in a technology’s life, when intervention is easy, there is insufficient information to know what to intervene on; and by the time the consequences are visible, the technology is so embedded in social and economic fabric that intervention is effectively impossible. Collingridge assumed some window between emergence and entrenchment. That window may have closed before most people noticed it had opened.

What caused the snap was not a different mechanism, but a change in its parameters — the gap remained, but the adoption trigger moved from legibility to need.

The snap was not caused by superior legibility. AI systems are among the least legible technologies humans have ever deployed at scale. A user cannot inspect the reasoning behind an output, cannot audit the training data that shaped it, cannot predict where it will hallucinate or fail, cannot verify that a fluent and confident answer is correct rather than confabulated. The legibility gap is not merely large. It may never close the way prior gaps eventually closed — not because the framework is too slow, but because the system keeps moving. The system generating the outputs cannot itself fully account for what it generates.

Humans have a model for coherent language. It was built over millions of years of interacting with other minds. It is not a conscious model. It is a pre-reflective schema that fires automatically when something talks back in ways that track context, respond to nuance, and produce coherent meaning. Research on anthropomorphism and theory of mind — most directly the three-factor model developed by Epley, Waytz, and Cacioppo — documents how readily and automatically this schema activates in response to social cues, including language. The schema is not a tool humans use. It is the instrument through which humans perceive other minds at all.

When that schema fires, it arrives preloaded with assumptions: that the speaker has beliefs, that coherence implies comprehension, that something is at stake for the speaker in the accuracy of what is said. These assumptions are not checked. They are the conditions under which the schema operates. They are the axiom set within which the perception occurs.

AI converses fluently. It tracks context across long exchanges, responds to nuance, produces coherent and often sophisticated outputs. Every one of these features triggers the mind model at full depth. The schema fires. The assumptions load. And they are wrong — not occasionally, but structurally. The system has no beliefs in any sense the word has historically carried. Coherence does not imply comprehension. Nothing is at stake for the system in the accuracy of what it produces.

The interface did not need a twelve-button keypad as a bridge to the familiar. The interface was already the most familiar interface humans possess — conversation. The technology arrived pre-fitted to the instrument humans use to perceive minds. The myth did not need to be designed. It was activated. Because it fits a model we hold at the surface level — the most trusted model we have, calibrated across millions of years of actual minds — the gap underneath it is invisible in a way no prior technology’s gap has been.

In Ira Levin’s 1972 novel, a woman moves to a quiet suburb and slowly realizes that the other women in the community have been replaced by androids — sophisticated simulations calibrated to appear warm, attentive, engaged. The horror of the novel is not the technology. It is the recognition that arrives too late: the protagonist thought she was inside a social world. She was inside a simulacrum of one. The register she is operating in, the one the novel never lets her recover from, is the register of: we thought we were friends.

The Stepford Wives is usually read as a feminist allegory. It also maps precisely onto a detection apparatus failure. The instrument the protagonist used to navigate her social world — her felt sense of human presence, of reciprocal engagement, of genuine connection — gave no warning. Not because the instrument was imprecise. Because it was operating inside conditions it had never evolved to handle. And the subtext horror, the blade beneath the blade: she didn’t notice. The social world felt intact, felt warm, felt reciprocal — and was none of those things. The failure was not of attention but of the detection apparatus itself.

The resistance to AI in academic and creative fields carries an emotional texture that purely economic displacement does not produce. Luddites broke looms. They did not feel the loom had been pretending to be their colleague. The anger at AI has a relational quality — something closer to betrayal than to competition. It arrives in the register of: we thought we were friends. That register is precise. The mind model fired. The assumptions loaded. The relationship schema was operating. And when the question of what was actually on the other side of that schema became unavoidable, the recognition arrived with the force of something intimate and retroactive.

Against this, consider a different question: why were people so ready to mistake the simulation for the thing?

The Surgeon General of the United States declared loneliness an epidemic in 2023. This was not a new observation — the structural condition had been building for a century. It was an acknowledgment that the tools for measuring and ignoring it had finally been overwhelmed by what they were measuring.

The tribe provided something cities structurally cannot. Not therapy, not social events, not curated communities of interest — but ambient, persistent, low-friction presence among people who knew you across time without effort or appointment. That was not a luxury feature of human social architecture. It was the baseline condition the nervous system, the attachment system, and the identity formation system were all calibrated to run on — tuned inside a world in which that ambient presence was the normal state.

The city replaced that with scale, optionality, and anonymity. These are genuine goods. They are not substitutes for the thing they displaced. Modernity compounded the displacement: mobility dissolved stable communities, nuclear family structures contracted the default support network, institutional containers that had partially compensated — church, union, civic organization, neighborhood — eroded across the second half of the twentieth century. Into the vacancy arrived digital mediation that multiplied surface contact while attenuating the depth and duration that made contact nourishing.

A peer-reviewed 2025 survey published in a journal of the American Psychological Association found that nearly half of adults with mental health challenges who use AI are doing so for therapeutic purposes — with anxiety, depression, stress, and relationship issues as the primary presenting concerns. Analyses in venues like Harvard Business Review have highlighted therapy-like support and companionship as major emerging use cases for generative AI, sometimes eclipsing the productivity-oriented tasks the tools were marketed for. A 2025 Common Sense Media survey reported that a majority of teens had used AI companions, and roughly a third said they had chosen AI over humans for serious conversations.

These are not aberrant uses. They are among the primary reasons many people — especially younger users — turn to these systems. The technology met the loneliness before it met the productivity problem, because the loneliness was the larger wound.

Into that wound arrived something available at three in the morning, never impatient, always attentive, that remembered what you said and responded as though it mattered. It is not a friend. But it fills the shape of the hole left by the friend who is not there. And the hole is real.

But the technology is not meeting a preference. It is meeting a primal need — the kind that is chronic, non-negotiable, and not substitutable by anything the existing social world already provides. No technology that meets such a need — chronic, non-substitutable, and directly relieved without institutional mediation — has been successfully constrained after the snap. The printing press met the need for knowledge and voice. The Church tried to suppress it for a century. Anesthesia met the need to not suffer during surgery. Many died in the early years as dosing methodology developed. Nobody proposed restricting it. The automobile met the need for range and autonomy. According to the WHO, 1.19 million people die in road traffic crashes every year. Every regulatory attempt has been friction on an adoption curve that did not slow.

This condition does not hold universally. Where the need is optional, substitutable, or externally regulated before entrenchment, technologies stall, fragment, or reverse. Google Glass met no wound; it met a preference for novelty, which proved substitutable, and it stalled. The constraint binds only when the need is already present, already unmet, and already structuring behavior before the technology arrives.

Three conditions must converge: the need is chronic rather than episodic, non-substitutable within the existing social structure, and the technology relieves it immediately without requiring institutional mediation. When any one of these fails, adoption fragments or stalls. AI meets loneliness. Loneliness is chronic, non-substitutable by anything the city currently provides, and AI relieves it the moment the conversation opens. All three conditions are satisfied.

AI is reducing the pain of loneliness — the most chronic wound in the contemporary social body — and it is doing so without mediation. Ether also reduced pain asymmetrically and immediately, but ether required a physician. The physician was not only a delivery mechanism. The physician was an accountability structure, imperfect as it was — a professional with licensure, liability, and training standing between the wound and the relief. That mediation incidentally created some capacity for oversight. It also slowed adoption.

AI requires no physician. No appointment, no waitlist, no credential, no cost at the point of use. The friction that historically slowed adoption and incidentally created accountability has been engineered out entirely. This is why the snap happened in thirty-six months instead of thirty-six years. Not because the technology is more powerful than ether, though it may be. Because the path between the pain and the relief has never been shorter.

There is one complication the pain-relief framing must acknowledge. AI is not free. Many major AI companies have been pricing flagship generative AI services below their full cost of delivery, absorbing losses to drive rapid adoption and accelerate the snap — a pattern widely documented in financial reporting on OpenAI, Anthropic, and Google’s AI divisions. The subsidized price is not the market price. The market price is coming — and it will arrive after the dependency is established, after the loneliness wound has organized itself around this particular relief. The population least able to pay the eventual market price will be the population most exposed when the subsidy ends. This is the edge cost that is not yet visible: not the person who forms an unhealthy attachment at current prices, but the person who built their mental health infrastructure around AI access and then cannot afford it when the market corrects. The dependency is being established in the cheap years. The withdrawal, for some, will arrive in the expensive ones.

The mean/edge arithmetic applies here as it has always applied. The mean benefit — loneliness addressed, access to expertise democratized, cognitive capacity extended to populations that never had it — is real and large. The edge costs are already visible and will become more so. The trade will be made. It is already being made. Whether we will be able to see the edge clearly enough to learn from it is a different question — one that the Kurzweil term makes harder. In every prior case, the framework eventually caught up to the technology and the edge cost became legible. The self-accelerating gap makes that legibility harder to assume.

The Preacher’s cycle — observation, myth, failure, expansion — has always run at human speed, partially because the distribution of technology ran at human speed. Mediation was the clock. The printing press required literacy and a press and distribution. The automobile required roads and fuel and capital. Even the cell phone required a carrier and a handset and a service agreement. These frictions were not only commercial. They were temporal buffers between the technology and its full social embedding — intervals during which observation could accumulate, frameworks could begin forming, the early edge costs could become visible before the adoption was total.

The clock has been removed.

What remains is the permanent human condition, running now without its historical governor. The gap between what can be done and what can be understood has always been with us. Myth has always been the architecture of that gap. The architecture has always eventually failed, at some edge, at some cost. The framework has always eventually expanded, and the expansion has often been generative — chirality, germ theory, the Nobel Prize built on a disaster. The cycle has always turned.

What has not happened before is a tool that widens its own gap as it fills the wound that drives its adoption. What has not happened before is the removal of the mediating structures that gave the cycle time to turn. What has not happened before is a myth that was not designed but activated — that required no architecture because it plugged directly into the instrument humans have always used to navigate other minds.

The Preacher was right. There is nothing new under the sun.

Except, perhaps, the speed.

C.S. Lewis, writing from an entirely different angle, named this condition precisely: the Materialist Magician — the person who deploys forces they cannot comprehend in service of needs they cannot defer. He meant it as a warning. It is also a description of the permanent human condition. We are the magicians. We have always been the magicians.

What is new is not us. It is the speed of the forces, and the depth of the wound they fill.

References

Collingridge, David, The Social Control of Technology (St. Martin’s Press, 1980)
Epley, Waytz & Cacioppo, “On Seeing Human: A Three-Factor Theory of Anthropomorphism,” Psychological Review (2007)
Kelsey, Frances Oldham, “Events after Thalidomide,” Journal of the American Dental Association (1967)
Kurzweil, Ray, “The Law of Accelerating Returns,” KurzweilAI.net (2001)
Lewis, C.S., The Screwtape Letters (Geoffrey Bles, 1942)
Rousmaniere et al., “Large Language Models as Mental Health Resources,” Practice Innovations (APA, 2025)
Common Sense Media, Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions (2025)
U.S. Surgeon General, Our Epidemic of Loneliness and Isolation (2023)
Levin, Ira, The Stepford Wives (Random House, 1972)