1. Video Call
November 2025, Hai Phong. 9 p.m.
The dormitory room is narrow. A single bed, a plastic shelf, a calendar on the wall. Lee Jung-hoon sets his laptop on the bed and opens the screen. Hai Phong's nights are humid. Through the window, motorbike horns mix with the steam rising from pho stalls. Even with the air conditioning on, November in Hai Phong stays sticky.
On the wall calendar, next month's Korea visit is circled. December 19 to January 2. Dates timed to his daughter's winter break.
The screen comes to life. His daughter's room appears. The poster has changed again. The idol poster that hung there last month has been replaced by something new, but the screen is too small to make it out. She is wearing a cardigan over her school uniform. November in Incheon is already cold.
The room on the screen looks small. Whether it is because the room's layout changed after Lee Jung-hoon left, or because the screen makes it look that way, he cannot tell. On the desk, workbooks are stacked — mathematics, English, science. Three subjects piled to the same height, which means all three have been started and abandoned midway.
"Dad."
"Yeah."
A moment of silence. Fluorescent light falls across his daughter's face. A textbook lies open, a smartphone beside it. He tries to read her expression through the screen, but the image quality is not sharp enough.
She speaks.
"Dad, what should I be studying."
Lee Jung-hoon's hand goes still. That is not a question — it is a sentence. The question mark is missing. Her homeroom teacher has been doing university-counseling sessions, she says. When she reaches third year, she will have to choose between humanities and sciences.
Her friends have started attending coding academies. ₩1.2 million a month. His daughter knows that the monthly net profit from Lee Jung-hoon's chicken franchise was ₩1.5 million. She is beginning to feel the weight of numbers.
Lee Jung-hoon spent 28 years reading the condition of molds by the sound of the press. Those ears could catch the friction of 0.03 millimeters of uneven wear. But he did not hear the sound of the structure changing. Until the AI quality-prediction system was introduced, he believed that what he was best at would always hold its value.
Those who were most faithful to the system are pushed aside most deeply. Lee Jung-hoon was the precise example of that paradox.
What to say to her now. He could tell her to learn coding. He could tell her to aim for SKY — Seoul, Korea, and Yonsei, Korea's top three universities. But for three months, commuting to the Hai Phong factory every morning at 5:47 a.m., Lee Jung-hoon has been learning one thing. The path that looks most certain may be the most dangerous path.
Instead of an answer, he turns the question back.
"What is the one thing you understand least."
His daughter pauses. Lee Jung-hoon waits. Humidity collects on the window glass in Hai Phong. For this question to reach her, the satellite signal has traveled 36,000 kilometers up and back down. That distance is the time that lies between Lee Jung-hoon and his daughter.
She speaks. "Math. No, I don't know." She points at the three workbooks and gives a small smile. "I don't understand any of it."
That answer stays with him. Not in a bad way. The sound of the press, the tremor of the mold, the 0.03 millimeters of uneven wear — the things he believed he understood best were classified as "obsolete data" within three years. The honesty of "I don't understand any of it" may outlast the certainty of "I understand everything" — that is what Lee Jung-hoon has been learning since he arrived in Hai Phong.
But how do you convey that to a sixteen-year-old. Lee Jung-hoon looks at the screen and opens his mouth. Closes it. A rivulet of condensation runs down the window.
2. The Weight of the Numbers
₩800,000 flows to Lee Jung-hoon's daughter every month.
Total private education spending for Korean elementary, middle, and high school students in 2024 was ₩29.2 trillion. A fourth consecutive all-time high. Student numbers are falling while private tutoring costs keep rising. From roughly ₩21 trillion in 2019, the figure grew by ₩8 trillion in five years. 1.2 to 1.3 percent of GDP. Roughly half the size of the ₩59 trillion defense budget. Korea spends half of what it spends on military deterrence on its children's test scores.
Monthly private tutoring expenditure per student averages ₩474,000 nationwide. For high school students it is ₩700,000; middle school, ₩540,000; elementary school, ₩360,000. The ₩800,000 per month spent on Lee Jung-hoon's daughter is close to the Seoul middle-school average.
Lee Jung-hoon was behaving like "an average Korean parent" — pouring more than half of the ₩1.5 million monthly net profit from the chicken franchise into his daughter's education. Now that the franchise has closed, those ₩800,000 come from his wife's salary and his Hai Phong remittances.
When the cumulative scale of this investment is calculated, the outline comes into view. Roughly ₩26 million over six years of elementary school, roughly ₩19 million over three years of middle school, roughly ₩25 million over three years of high school. Private tutoring alone: roughly ₩70 million. Adding four years of university tuition and job-preparation costs, the lower bound is ₩90 million and the upper bound ₩300 million. That is the total investment required to put one child on Korea's "standard path."
That figure corresponds to the deposit on a small apartment on the outskirts of Seoul. Korean parents are putting the price of a home into their children's heads.
Among OECD member countries, Korea's share of private funding for higher education is 37.2 percent. The OECD average is 16 percent. Finland is 2 to 4 percent. Korean parents pay more than double the OECD average out of their own pockets. The state's obligation for education's insurance function has been transferred to the individual. What should be a public responsibility is paid from private wallets.
This is why private tutoring spending behaves less like a rational choice and more like structural coercion.
Korea's total fertility rate in 2023: 0.72. The lowest in the world. The fact that private tutoring costs set a fourth consecutive record while fertility hits the floor is not a paradox. It is a perfectly logical outcome.
Have fewer children and concentrate investment on one. You can spend ₩300 million on one, but it is hard to spend ₩100 million each on three. Low fertility and high tutoring costs are two faces of the same decision. The fact that Lee Jung-hoon's daughter is an only child compresses this structure into one household.
Lee Jung-hoon is betting everything. He does not yet know precisely where that "everything" is headed.
Follow the direction in which ₩29.2 trillion flows — and you arrive at a map of AI exposure.
3. The Fragility of the Destination
The ultimate destination that ₩29 trillion is aimed at is: white-collar positions at large conglomerates, finance, civil service, and the professions. Tracking the main employment paths of SKY graduates, four directions stand out clearly. Corporate planning, strategy, and finance: 40 to 50 percent. Financial sector: 10 to 15 percent. Civil service: 8 to 12 percent. The professions: 10 to 15 percent. And these four directions are precisely the ones most exposed to AI.
Combining the LLM-exposure analysis published by Eloundou et al. in 2023 with the Goldman Sachs report of the same year, a map emerges that runs counter to intuition. 46 percent of office and administrative work carries the potential for AI replacement. Legal services: 44 percent. Finance: 43 percent. Construction, by contrast, sits at 6 percent; facilities maintenance at 4 percent.
The higher the credentials, the more polished the academic-credential hierarchy — the more vulnerable to AI. The mold-assessment ability Lee Jung-hoon built over 28 years was replaced by AI, but what he does in Hai Phong — teaching people on the factory floor and sensing the condition of equipment through his body — has not yet been replaced. Because the work requires not factory data but factory context.
Why is that so. Three structural reasons.
First, occupations whose core is language processing are the direct target of large language models. Writing reports, reviewing contracts, composing analytical memos — these are the core tasks of SKY white-collar workers, and also what LLMs have been best at since GPT-4.
Second, professional work with standardized procedures is easier to automate than non-standard manual labor. A plumber's hand skills are hard for a robot to replicate; a lawyer's review of a standard contract is easy for an LLM to replicate.
Third, fields with abundant digitized data — financial statements, case law, administrative documents — are automated first. Work that requires touching, smelling, and judging by sense has not been digitized, and so remains in AI's blind spot.
Here a critical conceptual distinction is needed. AI is not eliminating occupations. It is decomposing the tasks within occupations. The logic we analyzed in Chapter 4 of this book translates here into a question of educational investment.
Whole law firms are not disappearing. Instead, the contract-review work that junior lawyers once performed is being replaced by AI, and the hiring of junior lawyers is declining. Bank branches close; AI assessment systems are introduced; lower-level examiners are reduced. The entry pathway is being sealed off.
The top of the profession remains, but the bottom rungs of the ladder are being dismantled. Demand for partner-level lawyers' high-end counsel persists, but the experience-accumulation path to partnership — the process by which the repetitive contract review of the junior years builds a sense for legal risk — disappears.
The trend in the college wage premium shows this structure in numbers. In the early 2000s, the ratio of college graduates' wages to high-school graduates' wages was 150 to 160 percent. Between 2022 and 2023 it contracted to 130 to 140 percent. Translated into monthly wage differences: roughly ₩800,000 to ₩1 million.
Dividing the total educational investment of roughly ₩150 million by the annual wage difference of roughly ₩10.5 million gives a payback period of roughly 14.3 years. Fourteen years after graduation puts a person at 36 to 37 years old. If AI begins decomposing the tasks of that occupation at that point, the payback either shortens or becomes impossible.
Breaking down the expected returns on private tutoring investment into probability paths makes the numbers sharper still. The probability that concentrated private tutoring investment translates into SKY admission: 4 to 7 percent. The probability that SKY admission translates into employment at a large conglomerate or in the professions: 50 to 60 percent. The probability of remaining employed at a large conglomerate for ten years or more: 50 to 60 percent.
The probability of the full success path: 1 to 2.5 percent. For Lee Jung-hoon's daughter: ₩800,000 a month, over six years, roughly ₩57.6 million. The probability of success on the path that investment is targeting: 1 to 2.5 percent. And the occupational categories at the end of that path are at the front line of AI task decomposition.
That Google, Apple, and IBM have begun eliminating their four-year degree requirements is an early signal that the signaling function of education is weakening. Lee Jung-hoon does not know these numbers. But from the experience of having his career classified as "obsolete data" within three years, he has the same intuition the numbers convey.
4. Four Mirrors — Finland, Singapore, Israel, Germany
If the ₩29 trillion investment structure does not work, what does? The educational experiments of four countries serve as mirrors.
Finland teaches less and makes students think more. Following the sweeping reform of the national core curriculum in 2016, phenomenon-based learning was formalized. Instead of learning equations in mathematics class, students explore the phenomenon of "climate change." In that process, they draw maps of carbon emissions in geography, understand the properties of carbon dioxide in chemistry, analyze the carbon tax mechanism in economics, and read climate agreement documents in language class. The boundaries between subjects dissolve in the face of the phenomenon.
Among the seven core competencies, "multi-literacy" sits at the center. Reading text, reading data, reading images, reading code, reading social situations — these are integrated into a single ability. The ability to read structure is the central goal.
Teacher selection competitive ratio: 10 to 1. A master's degree required. No standardized tests. Finnish students' daily class time is 4 to 5 hours. Homework is at the OECD minimum. PISA 2022 science score: 511, above the OECD average of 485. The paradox of teaching less while producing better results lies in the fact that the unit of what is being taught is different.
The most underestimated element of Finnish education is the social status of teachers. A social compact exists in which teachers are treated as professionals. Teaching without a textbook is permitted; there is no supervisory inspection. That teachers can design the curriculum autonomously means society trusts them to exercise professional judgment. Without this condition, importing the methodology alone does not work.
Singapore extended the time horizon of education to the entire lifespan. SkillsFuture, launched in 2015, provides every citizen aged 25 and above with a learning credit of 500 Singapore dollars. For those aged 40 and above, up to 90 percent of education costs are subsidized.
The Skills Framework, which maps "skills needed now and skills needed in five years" across more than 30 industries, determines the direction of training. Demand leads supply. This is the reverse of Korea, where training institutions design courses and recruit students.
Adult learning participation rose from 35 percent in 2015 to 51 percent in 2022. The rate of increase in learning participation among those aged 40 to 60 was higher than among young people. In a city-state where, as Lee Kuan Yew put it, "the only thing we have is people," the continuous renewal of human capital is not welfare — it is a survival strategy.
That Lee Jung-hoon, at 53, is learning again in Hai Phong is an individual decision. In Singapore, it is a national system.
Israel puts its 18-year-olds to work on real-world problem-solving. Unit 8200 is a signals intelligence unit equivalent to America's NSA. The top roughly 2 percent, selected in high school, learn analysis, coding, systems design, and teamwork simultaneously in the field rather than the classroom. Situations requiring fast decisions with incomplete information. A culture that tolerates failure. A structure with low hierarchy and active lateral communication.
The people who emerged from this pipeline built Check Point, Waze, and Mobileye. The number of Nasdaq-listed companies from Israel, with a population of 9.3 million, is the highest in the world per capita.
The difference between a culture that asks "What question did you ask at school today?" and one that asks "What score did you get?" is not a matter of cultural taste. In an era when AI stores all "known answers," the value of "excellence within the known range" declines. What AI cannot answer is the questions for which there are not yet any answers.
Germany does not separate learning from work. In the dual system of vocational education (Duale Ausbildung), trainees spend three to four days a week doing practical work at companies and one to two days a week studying theory at vocational schools. 325 accredited training occupations; 490,000 training contracts as of 2023. Companies invest roughly €17 billion annually in training costs.
Pay is provided during the training period. €600 to €1,200 per month. Learning and earning simultaneously. Youth unemployment: 5.9 percent, among the lowest in the OECD.
After completing the dual system, obtaining a Meister qualification allows graduates to start a business, and some are given university admission eligibility. No social stigma. Germany's university enrollment rate of 33 percent contrasts sharply with Korea's 76 percent, but in a society where not going to university is a choice rather than failure, credential inflation does not occur.
The response to the AI era has also been swift. Since 2020, roughly 100 occupational profiles have been revised to include digital technology content. Mechatronics trainees receive alerts from AI predictive maintenance systems, inspect actual equipment, and learn in the field how to read AI output. What Lee Jung-hoon did at the Asan plant — reading the sounds and vibrations of equipment — an 18-year-old German trainee is now learning alongside AI.
The surface of these four countries is different. But there is a direction they all share in the AI era. Teach competencies rather than knowledge. Have students experience practice rather than tests. Design learning not to stop at school age but to extend across the entire lifespan. Korea's ₩29 trillion in educational investment is not aligned with any of these three directions.
5. The Crack in the Returns Equation
The human capital returns equation that Jacob Mincer proposed in 1974 is the standard tool for measuring the economic value of educational investment. The wage return on one additional year of education is estimated at 7 to 12 percent. For half a century, labor economists around the world have used the Mincer equation.
This equation measures "what price the market has been placing on education up to the present." It is a figure based on past data. There are two reasons why this historical return cannot be assumed to hold in the future.
First, the returns in the Mincer equation rest on the assumption that education raises the ability to perform specific tasks. When AI begins performing those tasks directly, the wage premium for task-performance ability declines. Legal research, financial-statement analysis, and drafting report summaries are the core tasks that university education teaches — and simultaneously the tasks that LLMs are rapidly absorbing.
The Mincer equation assumed that technology complements human labor. When technology begins substituting for human labor, the premise of the equation collapses.
Second, the wage premium of a SKY degree comes from both the ability signal and the school effect. When employers can use AI tools to assess job fit directly, the signaling function of the degree weakens. Companies including Google, Apple, and IBM have already begun eliminating their four-year degree requirements. Portfolio, not diploma, is becoming the signal.
Credentials are a ranking within the current system. When the system changes, the ranking changes. The ability to read is the ability to create a new ranking when the system changes.
This is what Lee Jung-hoon's 28 years proved. In 1996, when he graduated from Inha University's Department of Mechanical Engineering and joined Hyundai Motor, his return on educational investment sat within the range the Mincer equation predicted. His trajectory — rising to senior department head of production technology at a large conglomerate — was a success story of educational investment.
But when the AI quality-prediction system was introduced in 2023, three years consumed the accumulated returns. The equation itself had changed. The Mincer equation works when the technological regime is stable. At the moment the technological regime shifts, past returns do not predict future returns.
The paradox of overqualification that we already analyzed in Book 5, Chapter 14, The Strategy of the In-Between, translates here into the language of education. Korea's university enrollment rate of roughly 76 percent is among the highest in the OECD. Compared to Germany's 33 percent, it is more than double. Yet the unemployment rate for university graduates — roughly 6.7 percent — is higher than that for high-school graduates — roughly 5.5 percent.
This paradox of higher credentials correlating with higher unemployment is not a statistical error — it is a structural outcome. Three forces overlap: the aspiration effect, in which university graduates target only positions they consider commensurate with their credentials; oversupply, in which the number of graduates universities produce each year exceeds the rate at which high-credential jobs are created; and the winner-takes-all structure, in which large conglomerates use educational credentials as a screening tool. Korea has roughly 190 four-year universities and roughly 130 junior colleges — one of the countries with the highest number of universities per capita in the OECD.
Lee Jung-hoon's daughter was born in 2010. Accounting for life expectancy, she will live to somewhere between 2090 and 2095. Even assuming retirement at 65, she will have 25 to 30 years after retirement.
The three-stage model — education to employment to retirement — does not work for this generation. The design of completing education at school age and living the rest of one's life on the returns of that education collapses in front of a working life exceeding 60 years. Lee Jung-hoon's experience was not sufficient. His daughter's 60 years cannot be sustained on a single education even less. For this generation, multi-stage design is not a choice but a necessity.
6. Coal Mines and Cram Schools — The Blind Spot of Good Intentions
1833, Yorkshire, England.
A nine-year-old girl pulls a coal cart in a mine shaft. She crawls on all fours through the dark. Fourteen hours a day. The shaft ceiling: 60 centimeters. Spaces adults cannot enter. Her value lies in the smallness of her body.
The reason her parents sent her to the mine is simple. She could earn money. Mine labor was a realistic choice. To keep the family fed, so long as the child's body could bear it, and so long as the mine needed coal, this was the most certain path available.
There is no malice in the parents' intentions. Only good will and practical judgment. "Working here means we can survive." Making the best choice within a currently operating system.
2025, Daechi-dong, Seoul.
A sixteen-year-old student moves through three cram schools a day and receives four hours of private tutoring. Seven hours at school, four hours at cram schools, two hours of self-study. Thirteen hours a day spent studying. The reason the parents send this child to cram schools is equally simple. Getting into SKY means a good job.
Private cram schooling is a realistic choice. "Studying here means getting into a large conglomerate." Making the best choice within a currently operating system.
The parents in both chairs have something in common. Good intentions, trust in the currently operating system, and the inability to read the signals of systemic change.
There is also a difference. The harm inflicted on the mine child was immediate and physical. Her back bent. Her lungs were destroyed. The problem was visible, and the 1833 Factory Act was eventually passed. Even that Factory Act, as we confirmed in Book 4, was not enforced until four independent inspectors had been deployed. A law without enforcement is just paper.
The harm inflicted by excessive private cram schooling is delayed and structural. When the child who received ₩90 million to ₩300 million of investment faces AI task decomposition in her early thirties, that harm first becomes visible. Because the time gap between investment and harm is long, recognizing the causal relationship is difficult.
The numbers we analyzed in Book 1, The Displaced and the Discerning, Chapter 10, echo here. The water-frame patent: 1769. The first effective Factory Act: 1833. Sixty-four years. During those 64 years, children in the mines kept crawling through the shafts. Compulsory education was realized in 1880 — 111 years after the technological starting point. Institutions always arrived late. But they never failed to arrive.
If we take 2022 as the AI starting point, we are only three years in. Just as the Factory Act was 64 years late, education law for the AI era will also be late. The question is how many children will be invested in paths that do not work during that period of delay. Lee Jung-hoon's daughter is one of those children.
The forces resisting educational reform come from four places. The private tutoring industry, which has a stake in ₩29.2 trillion; taxpayers who resist rising education costs; large conglomerates that use academic-credential hierarchy as a screening tool; and universities themselves, whose reason for existing is tied to the current system.
Reform is delayed where the interests of these four forces intersect. The five patterns of institutional delay we analyzed in Book 4 — information asymmetry, capture by vested interests, the cost of consensus, the gradualism of crisis, and ideological barriers — operate identically in education.
AI-driven task decomposition is a "gradual crisis." Every year the entry pathway narrows slightly, but each individual change is too small to appear in macro indicators. Without a triggering event, the policy window does not open.
The parents of the 1833 mines and the parents of 2025 Daechi-dong share the same good intentions. And the same blind spot.
7. The Ability to Read
Then what is to be taught?
Not credentials — the ability to read structure.
Structural literacy is the ability to grasp not the surface information visible before you but the rules and relationships of force that produce that information. Reading the fact that "AI is advancing" is fact-reading. Reading that "the tasks replaced first by AI's advance are digitized and repetitive, that occupations with many such tasks are at risk, and whether my own path falls into that category" — that is structure-reading.
"Korea's private tutoring spending is ₩29.2 trillion" is a fact. Asking "whether the occupations that investment is targeting are having their tasks decomposed by AI even as the investment scale keeps rising, and whether this is a result of information absence or alternative absence" is structure. The distance between fact and structure is the depth of literacy.
There are approaches extractable from the educational experiments of the four countries.
Finland's multilayered reading training makes students look at a single phenomenon through multiple lenses. The practice of reading the same number differently through an economics lens, a history lens, a sociology lens, a mathematical lens. "Why is this number this size." "When will this trend reverse." "Who made this rule." Forming the habit of questioning is the core.
Comparative analysis prompts the question of how the same phenomenon is handled differently in different countries. "Why does Korea have a 76 percent university enrollment rate and Germany 33 percent. Which is better. Under what conditions." Training in refusing binary choices. Asking not "is this right or wrong" but "under what conditions does this work."
Israel's deferred-judgment training builds the habit of not forming an immediate judgment when first encountering information. "If this news is true, who benefits. Who loses. Who made this information and why." The ability to analyze the source and motivation of information. Not consuming information that surfaces in a news feed, but asking why that information appeared on your screen right now.
Germany's and Israel's failure-analysis training dissects failure cases rather than success cases. "Why was Lee Jung-hoon displaced?" reveals the structure more sharply than "Why did someone succeed?" Read the pattern of failure and you can predict where that pattern will repeat next.
There are things AI currently cannot do. Reconstituting context in new ways, making trust judgments, deciding which goals to pursue, acting on incomplete information, negotiating with stakeholders. The common thread across these five is that all of them fall within the domain of "the ability to read." The ability to read structure, to read people, to read situations.
This ability does not begin with memorizing answers — it begins with generating questions.
Paradoxically, the most necessary education in the AI era is not "learning AI" but "learning what AI cannot replace." What the ₩1.2 million-per-month coding academy targets is the ability to build AI — but even that ability is being rapidly absorbed by AI. The speed at which GitHub Copilot and Cursor are replacing boilerplate code and simple debugging is faster than the speed at which coding academy curricula are updated.
This does not mean learning to code is meaningless. It means that learning only to code is dangerous.
The most valuable thing Lee Jung-hoon can give his daughter is not his severance package. It is 28 years of analyzed data. "Do you know why Dad was displaced? It wasn't because AI was faster and more accurate than Dad. It was because Dad couldn't read the structure changing ahead of time. I was running in the same direction, but the starting line had already moved." That sentence is failure analysis, and failure analysis is the most direct education in the ability to read.
The paradox of overqualification analyzed in Book 5, Chapter 14, The Strategy of the In-Between — the structure in which higher credentials correlate with higher unemployment — showed that accumulating more credentials is not the answer. The time has come for the eye that reads structure to replace credentials.
"The individual must read before the institution arrives." It took 64 years for the 1833 Factory Act to come. It took 111 years for compulsory education in 1880 to arrive. No one knows how long it will take for education law for the AI era to arrive. In the meantime, Lee Jung-hoon's daughter will already be in high school, will go to university or not, will find her first occupation or not. There is no time to wait for institutions.
8. The Threshold Question
We return to the video call.
His daughter speaks. "Why does Korea have to study this hard."
Lee Jung-hoon smiles. Since coming to Hai Phong, the number of times he smiles in front of his daughter has increased. It has been that way since returning from Park Sang-ho's gathering. Since "together" became possible. Lee Jung-hoon does not know that his daughter's question has touched the structure precisely. But he feels it is a good question.
"Good question. Dig into that."
He wants to say more. Every morning he went to work before dawn, assessed molds, brought down defect rates. All those hours left the factory without being translated into data.
In Korea, that experience was "obsolete." In Hai Phong, it was revived as "knowledge that still cannot be done without a person." When the position changed, the value changed. Yet he also knows this will not last forever. The "Smart Factory 2027" poster on the factory wall marks his expiration date.
In two years, AI could take his place in Hai Phong too. What Lee Jung-hoon needs to do until then is not to make himself irreplaceable — it is to have eyes that can read when that replacement is coming.
There are things he can say to his daughter and things he cannot.
What he can say. The reason Dad was displaced was not incompetence — it was that he could not hear the sound of the structure changing. Test scores are a ranking within the current system, and systems change. Reading what is changing matters more than knowing where to work.
What he cannot say. A certain answer about his daughter's future. Which major is safe, which occupation will survive, whether she should go to the coding academy or not. Lee Jung-hoon is a person whose career was displaced within three years. He has no certain answers. Claiming otherwise would be a lie.
But there is one thing he can give. Converting his own experience into failure data and handing it to her. An honest analysis of "why was I displaced." That is the most direct education in the ability to read.
The tacit knowledge Lee Jung-hoon accumulated at the Asan plant — the instinct of assessing molds by sound, the experience of reading the behavioral patterns of suppliers — was classified as "obsolete data" in Korea. But when that experience is analyzed through the frame of "why did this lead to this outcome," it becomes not obsolete but a lesson. Failure data is the most honest textbook.
The call ends. Lee Jung-hoon closes the laptop and looks out the window. Motorbike lights flow through Hai Phong's night streets. Far away, the harbor lights are visible. They overlap with the nighttime lights of the Asan plant. The same light, a different country, the same question.
The heat of the laptop fades. The screen that held his daughter's face a moment ago is dark.
The next chapter's story begins with Choi Eun-jeong (52). While Lee Jung-hoon is asking "what to teach," Choi Eun-jeong is asking "what becomes scarce."
If education is a question of designing the individual's future, the movement of scarcity is a question of redrawing the value map of society as a whole. If Part 3 examined the adaptation of individuals and collectives, Part 4 asks for the coordinates of where that adaptation is heading.
Lee Jung-hoon's daughter opens a textbook, but her father's question stays in her head. "What is the one thing you understand least." Lee Jung-hoon knows he cannot answer that question himself. What he can give is not an answer but the weight of a question. And he knows too that the way to bear that weight cannot be taught. If what you can hand down to the next generation is not the right answer — then what is it?