When Intelligence Is Cheap, Understanding Is Expensive

These days, it’s not uncommon to receive an email, watch a YouTube video, or read a blog post that has clearly been written by AI but isn’t the usual “slop.” It is unusually sharp—well-structured, perceptive, and full of connections that land just right. What we are encountering is the outward form of intelligence: fluent, articulate, and often genuinely insightful output. To be clear, this is not the same thing as deep personal understanding. While most people instinctively treat fluent intelligence as evidence of knowledge or truth, the two have always been separable—an insight sharpened by evolutionary psychology. AI has made this kind of fluent, articulate intelligence dramatically cheaper and more abundant, while the slower, more expensive work of understanding—testing, owning, integrating, and reality-checking that output—has not become cheaper at all, and is now becoming more valuable.

AI output clearly varies in quality. Not all of it is intelligent—hallucinations, fabrications, and shallow responses are still common. Yet the synthetic intelligence of large language models is advancing rapidly and becoming increasingly profound. High-quality intelligent content is poised to be everywhere, reshaping how we work, learn, communicate, and create.

Importantly, this shift is democratizing expression in powerful ways. There is a great deal of valuable human intelligence—ideas, observations, and hard-won perspectives—held by people who have never been particularly good at writing. For them (and for many of us), articulating thoughts has long been a significant hurdle, fraught with emotional friction. AI removes that barrier. It helps people express ideas they’ve held for years, often enabling deeper and clearer thinking than before. The process of writing no longer blocks the thinking.

I’m not opposed to this—far from it. Along with the inevitable slop, we’re about to be flooded with thoughtful, intelligent artifacts and worthwhile material, including from voices that previously struggled to be heard. The challenge isn’t that the output is inherently fake or worthless. It’s that high-quality intelligence has become dramatically inexpensive to produce.

For most of human history, fluent intelligence and genuine understanding were tightly coupled. Generating well-connected prose required real cognitive work, so articulateness served as a decent proxy for depth. We evolved to trust the signal.

AI broke that proxy. You can now ask AI to generate articulate arguments, insightful connections, and useful observations with almost no personal investment. What the AI produces is frequently intelligent and valuable. What it cannot do, however, is transmit the hard-won personal understanding that comes from wrestling with the ideas yourself—testing them against reality, revising under pressure, and integrating them into your own larger picture of the world.

Not all fluent human communication is accurate, either. People have always produced intelligent-sounding nonsense, motivated reasoning, or elegant misdirection. But the high cost of fluency acted as a natural filter. Now that filter is largely gone. The value, therefore, shifts decisively to what happens after the intelligence appears.

This is analogous to what’s happening in education. The old proxies for learning—completing assignments, turning in homework, producing fluent papers—have been hollowed out. When anyone can generate those artifacts instantly, what becomes precious is actual learning: the internal work of grappling with material, making it your own, and developing the capacity to use it wisely.

It’s a dramatic (if highly magnified) parallel with the shift from analog to digital photography. Digital tools obviated the need for deep mastery of film, light, exposure, and development, yet enabled far more people to create at a higher level. AI is doing the same for ideas.

The real work now moves to the human side: leveraging the output, comparing it with other perspectives, stress-testing it for hidden assumptions or weaknesses, and figuring out how (or whether) it fits into bigger pictures. Does this intelligent artifact meaningfully inform the topic? Does it hold up under scrutiny and real-world tests?

Early in the AI wave, a Claude advertisement captured the exciting potential: “Find your problem.” With this much intelligence at our fingertips, we can tackle challenges that once required years or decades of dedicated study. AIs can coalesce vast swaths of human knowledge, surfacing connections across domains that were previously almost impossible to see. Even though not all recorded knowledge is accurate, this accessibility creates fertile ground for genuine insight.

This is where the human element becomes critical—and potentially transformative. The AI supplies raw intelligence and connections. The human brings discernment: evaluating how pieces relate, weighing them against reality, spotting gaps or misdirection, and steering toward deeper understanding. Enough of that sustained, disciplined work compounds into wisdom no model can fully replicate. Done well, this partnership opens enormous possibilities for breakthroughs that no one could have achieved alone.

The same inversion applies here. The fluent, intelligent artifact is no longer rare or expensive, so it can’t reliably signal personal understanding. What has grown precious is the human endeavor that follows—the management, curation, thoughtful application, and integration of all this cheap intelligence. The true test is what survives real-time defense, experimentation, iteration, and honest scrutiny.

There is also an inward cost if we skip that step. Every time we let the machine do the heavy lifting without the subsequent human work, the mind that could have been strengthened by wrestling with the ideas stays underdeveloped. We keep the credit and lose the growth. We risk becoming riders narrating from scripts we didn’t fully author or internalize. The separated mind—fluent on the surface, less anchored underneath—finds its perfect technological companion. If history holds, this will be the outcome for most people. But not all.

We are entering a world where articulate, intelligent content will be everywhere. You will no longer be able to assume that such a piece has a fully present, deeply engaged mind behind it. The polished email, the insightful post, the compelling video—these are no longer reliable proxies for personal understanding. That’s the downside.

But when human understanding uses these intelligent tools for leverage, we are likely to find that incredible explorations of ourselves and the world are just beginning to take place.

The Functional Fiction Framework of Human Nature

One question has organized serious thought about human nature for as long as such thought has existed: Why does the persistent gap between what humans say about themselves and what they actually do remain so consistent across cultures, eras, languages, and registers? Philosophers, historians, sociologists, evolutionary psychologists, cognitive scientists, contemplatives, and novelists have each named pieces of it. None has produced a single integrated account that explains why the gap forms, why it takes the shapes it does, why it recurs at every scale of human organization, and what kinds of intervention can meaningfully reduce it.

This essay introduces a framework that does that work. It rests on evidence made possible by a methodological capacity that did not exist five years ago, paired with an architectural account of the human mind that explains what the evidence shows. From these emerge three practical principles that predict the patterns earlier traditions described but did not fully explain. The framework connects to major prior accounts of human nature without displacing them, integrating their observations into a unified picture. It also predicts its own reception, freeing the work from any need for immediate widespread acceptance.

A functional fiction is a durable pairing of an idealized narrative with the operative function it covers. The narrative is sincerely held and publicly defensible. The operative function is the actual movement of value the structure produces. The space between them is the narrative-operative gap. The framework’s central claim is that this gap is not a moral failing, curiosity, or artifact of bad institutions. It is the structurally inevitable output of an evolved architecture, replicated at every scale where separated-mind humans organize value.

I offer this as a testable hypothesis, not a proven theory—an integration that explains more of the human record than anything else I have encountered, and that has held up under the checks I could run. What follows is the evidence and reasoning, the three principles, the framework’s relationships to existing accounts, and a clear statement of its limits.

The Empirical Basis

In early 2026, I ran an identical prompt across multiple large language models (ChatGPT, Claude, Gemini, Grok, Qwen, DeepSeek, Manus). Each was asked to identify recurring patterns in human self-narration across its training data and to distinguish stated claims from what the structure of those claims revealed about actual motives and selection pressures. The models converged strongly on the fundamental architecture of human self-description, independent of their differing training.

ChatGPT captured it concisely: Human self-narration is consistently optimized to make competitive, status-sensitive, coalition-bound organisms appear morally governed, publicly oriented, and metaphysically justified.

This convergence matters: the pattern lives in the human record itself. Eight domains stood out where the narrative-operative gap appears most consistently: the hierarchy that must be denied, the altruism display, the innocence behind us, the enemy who completes us, the love that transcends, the gate called quality, the moral arc, and the sacred boundary. These are not random. They are precisely the areas where avowing the operative function carries the highest social cost, and thus generates the thickest narrative cover. Where honesty would cost the speaker, narrative thickens to protect value flow.

This is the first scaled view of human self-description across cultures, eras, languages, and registers—made possible by instruments that detect statistical patterns no individual reader could see. The methodology is reproducible by anyone with access to comparable models.

The Architectural Structure

The empirical pattern demands explanation. Why does this gap form so consistently, even across isolated populations, in forms specific enough for independent models to converge?

The answer is a three-layer architecture of the human mind.

The adapted mind—drawing on the work of Tooby and Cosmides—is species-wide firmware shaped by deep evolutionary selection. It handles survival, reproduction, threat detection, and the emotional machinery of social life. Fast, automatic, and largely unconscious, it communicates via feelings—dopamine, cortisol, oxytocin, fear, attraction, disgust.

The adaptive mind is cultural software acquired in development—my own extension of the model. It fits local language, kinship, religion, and economy into the firmware’s general capacities, allowing the same hardware to support a Yanomami warrior or a Manhattan banker.

The conscious deliberating mind—the rider—thinks, weighs, and speaks. It deliberates sincerely but operates with no direct access to the layers below. The first two layers (firmware + cultural software) form the elephant. The rider has no shared workspace with it. Deliberation is real, but the options, weights, and felt states are pre-shaped. (The rider-and-elephant metaphor has deep provenance preceding Haidt.)

This separation is architectural, not accidental. Intellect evolved primarily as a social organ—for reputation, alliance, status, and position—per the Social Brain, Machiavellian Intelligence, and argumentative theories of reason. Its relationship to objective truth is incidental.

Narrations therefore emerge optimized for keeping value moving, not for accuracy. They idealize because cultural templates reward alignment with ideals, and because honest narration of the elephant would often carry prohibitive social cost. The narrative-operative gap is what this architecture must produce.

The Selection Pressure on Pairings

The architecture ensures functional fictions will form. What determines which ones survive and elaborate is selection pressure on the pairing itself.

Variants compete. Operative functions are largely invisible to participants, so survival depends on narrative strength: how well it recruits engagement, sustains commitment, and resists scrutiny. Stronger pairings outcompete weaker ones, even with similar operative results. This is the Law of Inevitable Exploitation (L.I.E.)—a structural description of what selection rewards when extractive arrangements can persist, not a claim about conscious intent.

The modern school system illustrates it. Nineteenth-century variants competed; those pairing conformity production, custodial care, and credentialing with compelling narratives of empowerment and democratic opportunity survived and elaborated. The same logic applies to hospitals (healing narratives + billing), universities (transformation + rent), and religions (salvation + regulation). Narrative complexity grows over time because pairings inherit and refine prior cover under continued pressure. Stronger narratives correlate with higher avowal costs—the intensity clue in embryo.

Iatrogenesis follows predictably: systems whose stated purpose is help produce harm precisely because the narrative filters perception of costs.

The Selection Pressure on Individuals

Institutions select people whose architecture integrates sincere narrative belief with effective operative performance. Those who cannot hold the narrative or deliver results are filtered out. What rises is realmotiv alignment: the strategic substrate of survival and approval motives that stays below the rider’s awareness.

This explains sincere leaders in extractive systems. Insincerity at scale is detectable and costly; sincerity is stable because the architecture supports it. Insider testimony therefore reliably reproduces the narrative, not the function. Outsider observation of actual outputs over time is more reliable.

Reform from within is difficult for the same reason: successful insiders embody the selected integration. Narrative tweaks rarely touch the operative layer.

Coordinated Action Within Structural Conditions

Most extraction is structural—emergent from selection pressures. The framework also accommodates conscious coordination and conspiracy as predictable overlays once valuable gaps exist. Psychopaths and coordinated actors thrive in environments already covered by sincere narratives. Conspirators themselves operate under the same separated-mind architecture, narrating their actions idealistically.

The framework holds both layers without false dichotomies. It moves the moral question from intent to response once evidence of the gap appears: real sabotage (active suppression) versus self-sabotage (failure to examine when possible).

The Fractal Nature

Because the architecture is universal, the gap appears at every scale: individual motives, dyadic relationships, institutional missions, and civilizational founding stories. This fractal quality explains recurring civilizational cycles (Spengler, Toynbee, Strauss-Howe, etc.): the underlying separated-mind architecture remains unchanged.

The Architecture Without Architects

Institutions appear designed but are mostly selected. Like the vertebrate eye, complex functional structures emerge via selection on variants, not intentional architects. Surviving pairings are those that best balance compelling narrative with sustainable value extraction.

The American founding is a rare exception: deliberate structural constraints against capture, placed outside the system (separation of powers, etc.). Most constraints internal to a system get absorbed into its cover. Durable alignment requires external checks plus strong reality-feedback (bridges fall; patients die; markets clear).

The Three Principles That Constitute the Framework

1. Inevitability of Formation: Wherever separated-mind humans organize value, functional fictions will form. Strongest pairings link consequential operative functions with compelling idealizing narratives. This is structural, not contingent.

2. The Intensity Clue: Emotional intensity around a narrative signals the avowal cost and importance of the protected operative function. Dispassionate domains have small gaps; armored ones have large ones. It does not distinguish cooperation from extraction but reveals what is being protected.

3. Futility of Narrative-Only Change: The rider cannot dissolve the elephant. Narrative reforms get absorbed or rejected; operative functions persist. Effective interventions build external structural constraints assuming the architecture, sustained by reality-feedback. Alignment is rare, costly, and requires maintenance—the exceptions that prove the default.

How This Framework Relates to Existing Accounts of Human Nature

I am not a credentialed scholar in evolutionary psychology, cognitive science, or related fields. The comparisons here draw heavily on LLM-assisted synthesis of the literature and my own reading; they are offered humbly as orientation rather than authoritative critique.

The framework is not Kahneman-style dual-process theory (both System 1 and 2 live in the rider). It is not modular mind theory (which describes the elephant’s internals), nor Freudian/Jungian unconscious (this is ongoing architecture, not repressed content).

It builds on the rider-and-elephant metaphor (with provenance long preceding Haidt) but generalizes it: the separation operates across all domains, not just morality, and adds the cultural adaptive layer for precision.

The closest precedent is Robin Hanson and Kevin Simler’s The Elephant in the Brain. They rightly highlight hidden motives, functional self-deception, and institutions built around signaling for status, loyalty, and affiliation—often in roughly symmetric coalition games. The shared territory is substantial: sincere narratives covering operative functions.

Where my framework departs—and extends—is in emphasis. When reading their book, I repeatedly noted what felt like an underweighting of asymmetric value capture and “core profit motives” in institutional settings. The LLM-scaled corpus patterns showed not just mutual signaling but systematic extraction: institutions positioned to draw value from those they nominally serve, with flattering narratives for the extracted. My framework treats the pairing as the unit of selection and highlights how this produces the observed asymmetries and iatrogenic harms. Hanson and Simler provide a strong foundation on motives and signaling; this work builds on it by examining the cultural/institutional consequences of selection on those pairings at scale. The two accounts are complementary rather than contradictory.

The framework’s contribution is integrative: a three-layer architecture (with the adapted mind from Tooby and Cosmides and the adaptive mind as my extension), selection on pairings, fractal application, and empirical grounding that together predict the patterns we actually observe.

How I Arrived Here

The framework is the product of a long arc of looking. Fifteen years ago, after extensive reading in history, I concluded that the history of the world is largely a history of power and control—an empirical observation about what the record shows across cultures and centuries. The question of why the pattern was so consistent remained open.

In the years between, I worked on adjacent questions: the structural critique of education (Gatto, Illich, and direct observation), credentialing as social sorting, institutional gaps between stated and actual functions, and recurring patterns of capture. These were pieces of an unsolved puzzle.

When AI systems became capable of scaled pattern recognition, I had a new instrument. The 2026 cross-LLM experiment was an attempt to surface what humans deposit unintentionally in their writing. The convergence confirmed the premise.

The architectural account followed from explaining the empirical pattern. The fractal claim followed from seeing the same gap at every scale of organization. The diagnostic practice—my long-running “Conditions of Learning” exercise with educators—had already been surfacing operative functions versus narratives for two decades.

I am not a credentialed scientist or historian, but an philosophically-oriented platform-builder with decades of observing institutional gaps. This framework is what that looking, combined with the new instrument, has produced.

I do not expect wide acceptance, and the framework explains why. This is the Cassandra Paradox: accurate perception threatens group narratives or individual comfort, and the resisting architecture is the one described. Institutional gatekeepers, comfort-seeking audiences, and romantic reformers all have reasons to deflect it. The framework offers no hero or easy transformation—another structural limit on its appeal. It will likely spread, if at all, through small numbers of careful readers. That is the channel for which it is prepared. This is not despair but realism.

What the Framework Does Not Address

The framework does not prescribe what humans should do beyond building external structural constraints and engaging contemplative practices that reach the elephant. It does not specify which practices best engage the lower layers, which constraints are worth building in specific domains, or what constitutes human flourishing—those questions belong to moral, philosophical, and theological traditions, and to ongoing work in what I have called Evolutionary Therapy.

It does not resolve the Paleolithic Paradox: humans built for ancestral conditions now live inside modern abstractions that amplify the gap. Small-scale, locally legible structures often fit the architecture better, but the framework only describes the conditions that allow flourishing, not its content.

The framework does not exhaust human experience. Love, beauty, conscience, and meaning exist within it but exceed what it names. It is itself produced by a separated mind and carries the limits of any such account.

The Framework as Predictive Hypothesis

I offer the Functional Fiction Framework as a hypothesis with clear testable features. It predicts the narrative-operative gap as structurally inevitable, the intensity clue as diagnostic, the persistence of the gap across scales, the failure of narrative-only reforms, and the success (while they last) of external constraints backed by reality-feedback. These predictions are checkable and have held up under informal testing.

Challenges remain: explaining variance in civilizational longevity, smooth versus catastrophic transitions, and periods where capture was arrested. Refinement through application to specific domains will test and strengthen it.

What This Framework Is For

The framework is not cynicism. Operative functions and idealized narratives both accomplish real work—sustaining cooperation, communities, and meaning. Understanding the architecture does not destroy it any more than understanding a bridge destroys the bridge.

It offers a structural account of why human life has the shape it does, a practical diagnostic for reading specific cases, and a prescription focused on external constraints and reality-feedback rather than better narratives or better people. For those seeking deeper understanding, clearer diagnosis, or more effective intervention within the architecture we actually have, it provides a usable map.