Chapter 10: When Production Approaches Free, What Do Humans Still Need?
In the prologue, you read a passage attributed to Schopenhauer.
That passage was grammatically correct, appropriately worded, and structurally complete — it began by citing a philosopher to establish authority, then used an emotional reversal to create resonance, and closed with a seemingly profound statement. If it appeared in your social media feed, paired with a misty mountain landscape, you would probably pause and give it a second look.
That passage was generated by AI. The entire process took no more than ten seconds.
What this chapter sets out to do is dismantle that generator all the way to the bottom. And then ask a question that Part III has been converging toward but has not yet faced head-on: once AI has lifted education's covers, exposed the substance and void of individuals, and illuminated the dead knot of corporate culture — after the three layers of distortion form a closed loop, what do humans actually still need?
The Industrialisation of Tautologies
Return to that Schopenhauer passage.
Take apart its structure, and you will see four layers of formula at work.
The first layer is the citation. Schopenhauer's name provides intellectual authority. Most readers will not bother to check whether this is actually something Schopenhauer said, nor investigate the original context. The citation itself is a threshold; it tells you: this article has scholarship.
The second layer is the emotional reversal. "The end of suffering is not boredom, but a kind of clarity that has been polished by experience." This sounds like insight, but does it withstand scrutiny? What is "clarity polished by experience"? Under what conditions does it hold true? For whom? It answers none of these questions, but it mimics the feel of answering a question.
The third layer is the consolation. "Your anxiety right now is not because you are not good enough." This sentence is always true — because it applies to any anxious person, without needing to know who you are, what you are anxious about, or how complex your situation is. It is a tautology: a statement you cannot even be wrong about, at the cost of saying nothing at all.
The fourth layer is the closure. "The answer you have been searching for has never left." The function of this sentence is not to provide an answer, but to make you stop asking. If the answer has never left, then you do not need to keep looking — you just need to "feel." It swaps out the need for thought and substitutes emotional satisfaction.
Stack the four layers, and you have a bestseller-generating machine.
How low is this machine's technical threshold? Build a Gemini App in Google AI Studio, write an Express backend connecting to the API, put an input box on the front end — emotional keywords, philosopher selector, companionship-mode toggle. The entire system's code is under two hundred lines; deployment takes under thirty minutes. The text it produces is statistically "correct" in perpetuity — because it never risks saying anything that could be refuted.
This machine requires no life experience from the author, no research into Schopenhauer, no history of having been anxious, clear-headed, or polished by hardship. All it needs is an API key and a prompt template.
But what is truly worth asking is not how easily this machine can be built. It is a different question: why does its output have a market?
One possible answer: what people need may never have been knowledge, but the feeling of being understood.
The reason that Schopenhauer comfort text makes someone pause late at night is not because it says anything useful. It is because it mimics the shape of "someone understands you" — it knows you are anxious, it tells you it is not your fault, it implies the answer is already inside you. The entire formula's design goal is not to convey thought but to manufacture the illusion of resonance.
"The feeling of being understood" and "actually being understood" are two different things.
Actually being understood requires the other party to know who you are, what you have experienced, what your specific predicament is. This process requires time, dialogue, and the simultaneous presence of both consciousnesses. A tautology cannot accomplish this, because its design principle is to hit the target without needing to know who you are — what it hits is not you, but the emotional common denominator shared statistically by all anxious people.
But when people are vulnerable, they cannot tell the two apart. You will misremember "the feeling of being seen" as "having been told something." This is why such texts are always pushed late at night, always appearing at your most vulnerable moments — because once emotion rises, people stop verifying.
Chapter 7 dismantled anxiety arbitrage. Chapter 8 dismantled the prose illusion. Chapter 9 dismantled corporate rhetoric. The industrialisation of tautologies is the intersection of all three — it uses anxiety as fuel, prose as shell, rhetoric as structure, producing text that makes you believe you have been understood when in fact nothing has happened. AI has increased this machine's production capacity by several orders of magnitude. But the machine itself is far older than AI — humans have been fabricating depth with tautologies without needing any technology at all.
Then, in a world where tautologies can be industrially produced, what is truly scarce?
Five Layers of Scarcity
Lay out everything dismantled across the first three parts, and a structure emerges.
Tools are no longer scarce. AI can write articles, draw images, write code, make presentations, build websites. Output that a few years ago required a team can now be done by one person plus one API. But when everyone has the same blade, possessing a blade is no longer an advantage. What is scarce is no longer the blade, but direction — do you know what you want to cut, and why? Chapter 11 will fully develop this question, but mark it here: direction scarcity is the first layer.
Answers are no longer scarce. You can ask AI any question in three seconds and receive a grammatically correct, structurally complete, seemingly reasonable response. But when answers are at your fingertips, the truly difficult thing is no longer finding an answer, but judging which answer applies to your situation and which will hit a wall under your constraints. Judgement scarcity is the second layer — in the educational distortion dismantled in Chapter 7, the greatest cover was making people believe that "knowing the answer" equals "having judgement."
Information is no longer scarce. The volume of information online already far exceeds what any individual can absorb. But an explosion of information does not automatically make you understand the world. Understanding the world requires not more information, but a world model that can organise vast amounts of information into causal relationships — you know what causes what, what constrains what, what changes under what conditions. World-model scarcity is the third layer. The Ring 0 problem dismantled in Chapter 6 is an example: you can read every technical document on Windows on ARM, but without a world model encompassing the experience of three generations of platform failures, you cannot see the landmine.
Content is no longer scarce. AI can batch-generate articles, social media posts, analysis reports, academic paper summaries. But when content floods, the reader's problem is no longer "is there anything to read" but "can I trust what I am reading." Trust scarcity is the fourth layer. The experiment dismantled in Chapter 1 — the same sentence distorted beyond recognition after three layers of media translation — will only intensify in the AI era. When you cannot even tell whether a passage was written by a human or generated by a machine, trust becomes the most expensive currency.
Production is no longer scarce. Building an app, recording a course, publishing a book, opening an online store — the technical threshold and cost of all these production activities are approaching zero. But the easier production becomes, the more nakedly a question is exposed: why are you doing this? For whom? What problem are you solving? Meaning scarcity is the fifth layer, and the deepest. The bestseller-generating machine can produce a "seemingly profound" article in ten seconds, but it cannot answer "why write this article." That "why" is not inside the API, not inside the prompt template, not inside any tool. It is inside the experiences a person has lived through, the choices they have made, the consequences they have borne.
The five layers stacked together point in the same direction: as each link of production approaches free, what is truly scarce shifts from the external to the internal — from tools, answers, information, content, and productivity, things that can be replicated, to direction, judgement, world models, trust, and meaning, things that cannot.
The former is output. The latter is consciousness.
But "consciousness" is too large a word; everyone pictures something different when they hear it. A more precise framework is needed to take it apart.
The tradition of the I Ching contains an ancient set of concepts: word (言), image (象), and intent (意). "Word" is language and text — the most surface layer, directly transmittable and replicable. "Image" is imagery, structure, pattern — one layer deeper than language, requiring observation and induction to discern. "Intent" is intention, judgement, direction — the deepest layer, the system of trade-offs that precipitates after a person has lived, hurt, and thought. Chapter 12 will fully develop this framework. Here, use it to locate the five layers of scarcity.
AI is nearly invincible at the level of "word" — grammar, structure, format, phrasing, all batch-generatable. The Schopenhauer passage produced by the bestseller generator is pure "word." At the level of "image," AI can imitate the shape — it can generate metaphors, tell stories, create a sense of vivid imagery, but the images it assembles are drawn from a pattern library, not grown from a specific "intent." What AI currently cannot touch is "intent" — because intent is not information, not pattern, not the statistical optimum. It is a person's deep understanding of the world, formed after having experienced enough: what is important, what is not, what is worth bearing costs for.
The first two layers of scarcity — direction and judgement — belong to the domain of "intent." The third layer — world model — lies at the boundary of "image" and "intent." The last two layers — trust and meaning — are entirely the territory of "intent." AI's free production capacity covers "word," is encroaching into "image," but remains at a structural distance from "intent."
This is why, after production approaches free, scarcity shifts from the external to the internal. Tools solve the problem of "word." The problem humans face is that of "intent."
Networked Thinking and the Consciousness Database
If the core of scarcity lies in "intent," a natural question follows: can intent be stored, organised, and retrieved?
On the surface, this is what knowledge management tools do. Notion, Obsidian, various wikis, various databases — they let you file and categorise articles you have read, notes you have written, thoughts you have had. But the underlying logic of these tools is tree-shaped: folders, tags, categories, hierarchies. An article belongs to the "AI" category, or the "Education" category, or both. Once categorised, the relationships between articles are static — they are filed in the same folder, but the system does not know what causal connections exist between them.
What is truly valuable is not the articles themselves, but the relationships between them — the structural similarity between a certain educational observation and a certain corporate case study, the causal chain between a technical judgement and a personal experience, the verification relationship between a thought jotted down three years ago and new evidence that suddenly appears today. These relationships are not tree-shaped; they are networked. They intersect, overlap, mutually correct, and mutually validate.
A knowledge base stores articles. A consciousness database stores the transformation patterns between thoughts.
This distinction sounds abstract, but it has a very concrete operational implication.
A note in a knowledge base tags "what this article is about." A note in a consciousness database also tags "how this concept has evolved across different points in time and different levels of cognition." The same concept — say, "tautology" — what you understood the first time you encountered it, what you understood after analysing several cases with it three months later, what you understood after placing it inside a larger framework (like word–image–intent) a year later. These three versions are not replacements for one another; they are three cross-sections of the same cognitive trajectory.
If your system can record a single concept's evolution across different cognitive levels — you might call it a Multi Version Tag, or whatever name you find fitting — you possess something more precious than knowledge itself: your own cognitive evolution history. You do not merely know "what you understand now"; you also know "how you arrived here, step by step." This path itself is part of your world model, and it is entirely personal — no two people will trace the same cognitive trajectory.
But what does a consciousness database actually look like? Not a tidy library. More like a chaotic film set.
A person whose thinking is driven by consciousness often operates not in words but in images — and not a single still image, but a dozen films screening simultaneously. A residual image from a dream, a technical architecture diagram read three years ago, the expression on a CEO's face in yesterday's news, a judgement that floated up in the middle of the night, a somatic memory from a turning point in a personal experience — all of these churn simultaneously in consciousness, more chaotic, faster, and denser than visual thinking alone.
If you write these things down, they look extremely fragmented: dream records, diary fragments, scattered technical notes, analysis of a news item, half-finished ideas. Any outsider who opens this database sees a pile of shards. No structure, no chapters, no narrative arc.
But the shards are connected. It is just that these connections exist in the consciousness of the person who wrote them, not on the surface of the text.
What AI does here is a very particular kind of collaboration. It can help you tag these fragments — not traditional classificatory tags, but consciousness tags: this dream record and that technical analysis share structural similarity; this diary entry from three years ago and this judgement written today have a causal link. AI can recognise the shapes of these connections, partially externalising the networked structure that previously existed only in your mind into a visible graph.
But externalisation is not transformation.
Fragments plus tags plus connections are still just a network of fragments. It is not yet a book. It is not yet text that another person can read. Between the consciousness database and readable writing lies a transformation step — you must compress the dozen films screening simultaneously into a single linear narrative, allowing a reader who has never seen any of those films to follow your thinking from beginning to end.
This step, AI can accelerate, but it cannot replace. Because the direction of compression — what to say first and what later, which details to keep and which to discard, what structure to use to carry what argument — these decisions all come from "intent." AI is the accelerator of transformation, but the source of "intent" is always the person.
This book is one product of that transformation step. Its fuel — those scattered images, dreams, diary entries, technical analyses, industry observations, personal-experience fragments — has long been in the consciousness database, already tagged, linked, and repeatedly validated. But transforming them into the text you are reading now still requires time. Not AI's computation time, but a person's digestion time — you must steep yourself in those fragments repeatedly until the direction of compression emerges on its own from the chaos.
There are many more books waiting to be written. The fuel is ready. All that is lacking is the time for transformation.
AI can help you organise articles, summarise, and build connections. But what it cannot do is judge for you: is this new piece of information saying the same thing as a thought you had three years ago? That judgement requires your own consciousness — you must remember what you thought three years ago, you must recognise two concepts that look different on the surface but are structurally similar, and you must place them together for reinterpretation.
This is why a "personal database" is not merely an efficiency tool but an extraordinarily important piece of infrastructure for the future world. It is your cleanest information source — knowledge that has been screened, verified, and organised by you, unpolluted by algorithmic recommendations and media agendas. But more importantly, if you build it as a consciousness database rather than merely a knowledge base, it becomes the external extension of your world model — a mirror that lets you see your own cognitive evolution trajectory.
Are AI's Limits Permanent?
At this point, an honest acknowledgement is needed.
The analysis above carries an implicit premise: AI can achieve word and image, but cannot achieve intent. This premise holds today. But is it permanent?
Every generation has said "AI will never be able to do X," and then been proven wrong.
In the 1990s, people said AI would never beat humans at Go. In 2016, AlphaGo proved them wrong. In the 2010s, people said AI would never understand natural language. In 2022, ChatGPT proved them wrong — at least at the surface level of understanding. Today we say AI lacks a world model, lacks genuine memory, lacks the capacity for self-reflection, lacks the ability to compress experience into judgement. How confident can we be that, ten years from now, these statements will not become another set of expired "AI will never"s?
The research community is already moving. World Models are an active research direction — making AI not merely do language prediction, but build an internal model of how the world works. On memory, long-term and personalised memory are being integrated into large language models. Agent frameworks enable AI to decompose tasks, invoke tools, and execute multi-step operations in real environments. Self-reflection enables models to evaluate their own output and identify their own errors. If any one of these directions achieves a breakthrough, the boundary of "AI cannot" drawn today will recede one step.
So an honest assessment should be: AI currently cannot touch the core of consciousness. But the word "currently" cannot be omitted.
However, even placing this uncertainty fully on the table, one thing remains unchanged.
Even if AI one day truly possesses a world model, possesses long-term memory, possesses the capacity for self-reflection — it still cannot answer one question: Why do this?
"Why" is not a cognitive problem. It is a choice problem. Cognition tells you what the world looks like. Choice tells you what you will do in this world. Cognition can be improved by more powerful computation and richer data. Choice cannot — because the premise of choice is "what you care about," and "caring" comes from the experiences a person has lived through, the pain they have endured, the values they are willing to bear consequences for.
An article can be written by AI. But "why write this article" cannot be answered by AI. A system can be designed by AI. But "why build this system" cannot be decided by AI. A direction can be analysed for pros and cons by AI. But "whether to take this direction" cannot be chosen for you by AI.
Articles are just output. Systems are just tools. Directions are just possibilities. What truly determines everything is the person making the choice — where their world model comes from, how many failures have calibrated their judgement, and what they are willing to pay an irreversible price for.
This is the question deeper than "can AI or can't AI."
The Convergence of Three Threads
Pull Part III's three threads together.
Chapter 7 asked: why can education not recognise ability? Because the measurement system measures "word" — exam scores, operational skills, standardisable output. The judgement and cognitive depth that truly matter are invisible in the measurement system's language.
Chapter 8 asked: what does AI expose? It exposes that prose is not content, code is not an asset, frameworks are not thinking. When production costs reach zero, only one naked question remains: do you actually have something to express?
Chapter 9 asked: why can corporate culture not be fixed? Because consensus costs, discourse-power battles, post-hoc attribution, and institutional misalignment of responsibility are not technology problems — they are power problems. AI can accelerate building, but it cannot accelerate consensus.
The three threads converge at the same point: every link of production is approaching free, but production itself was never the truly scarce thing. Direction, judgement, world models, trust, meaning — these things that come from consciousness — are.
And consciousness cannot be replicated by a tool, cannot be sold through a course, cannot be proven by a certificate, cannot be tracked by a KPI. It comes from the days a person has lived, the problems they have thought about, the pressures they have endured, the choices they have made. It is not inside tokens, not inside API keys, not inside anyone's subscription plan.
After production approaches free, what do humans have left? The answer points to one word — consciousness. But what is consciousness, really? Not the consciousness of philosophy textbooks, not the "does the machine have feelings" debate of the AI industry. It is something more plain.