A system built for scarcity
School was designed for a world in which knowledge was scarce. Books were limited, experts were distant, explanations were expensive and access to a good teacher depended on geography, wealth and social class.
Knowledge therefore had to be transferred from one human mind to another. Children attended school to acquire information they could not obtain elsewhere and store it for future use.
That world is disappearing.
A child can now converse with a system capable of explaining quantum mechanics at ten different levels, translating almost any text, generating counterarguments, writing software, analysing images and adapting its explanations indefinitely.
Knowledge is no longer scarce.
Explanation is becoming infinite.
Yet we continue sending children into buildings organised around the scheduled transmission of information—as if knowledge still had to be physically transported from one human head to another.
We are training children to become what machines are learning to be.
In Economic Scenarios for Transformative AI ↗, the Anthropic Institute explores how increasingly capable AI could reshape the American economy by 2030.
These are scenarios, not predictions. But their implications are difficult to ignore.
Even the median expectations collected from American adults resemble Anthropic’s “substantial change” scenario: significantly greater economic output accompanied by declining cognitive employment.
Our education systems devote fifteen or twenty years to preparing people for cognitive labour precisely when cognitive labour is becoming automatable.
If knowledge is everywhere, why learn anything?
Artificial intelligence can retrieve, combine and explain more information than any individual could acquire in a lifetime. Teaching children to compete with machines through the volume of information they remember is therefore pointless.
The objective can no longer be to possess the answer.
It must be to know what to do when an answer appears.
A person without knowledge can ask AI a question. But that person cannot reliably determine whether the question is badly framed, whether the answer is plausible, what assumptions it contains, what it omits or whose interests it serves.
Children still need to understand number, probability, causality, evidence, language, biology, history, power and human behaviour—not as a warehouse of answers, but as a cognitive immune system.
The purpose of knowledge after AI is not retrieval. It is orientation. The educated person of the future will not be the person who knows the most. It will be the person who can remain intellectually sovereign while consulting something that knows more.
The work improves.
The mind does not.
AI allows students to produce work that exceeds their actual understanding. A child may submit a sophisticated essay without becoming a better writer, solve an advanced problem without understanding its structure, or generate an argument without confronting the uncertainty from which genuine thinking emerges.
Schools will be tempted to treat this as a cheating problem. It is much more serious. It is a developmental problem.
If AI continuously removes intellectual friction, children may become highly capable of producing outcomes while losing the ability to originate, evaluate or defend them.
Before children delegate thought, they must experience what it means to think.
Eight years of school.
A lifetime of education.
Children may still need twelve or more years of protected education. But they probably need only eight years of conventional school.
- 03—07
Inhabit reality
Play, language, movement, music, stories, nature, making things and learning to coexist. Very little AI. Children must encounter reality before encountering its simulation.
- 07—12
Build the internal model
Strong foundations in reading, writing, mathematics, science, history and the arts. Not to accumulate information, but to construct the mental models required to question it.
- 12—15
Confront complexity
Projects, experiments, debate, collaboration and interdisciplinary problems. AI used openly and critically. Assessment of reasoning and judgment, not merely output.
- 15→
Leave school. Continue education.
A fluid ecosystem of advanced study, apprenticeships, research, entrepreneurship, artistic practice, community service and real-world work.
From lessons to missions
The industrial model places children of the same age in the same room, studying the same material, at the same speed, during the same hours, for more than a decade. Its fundamental unit is the lesson.
The new model’s fundamental unit should be the mission.
A mission might require students to restore a damaged ecosystem, construct a working machine, investigate a historical controversy, develop a local business, stage a play, care for vulnerable people or debate a question for which neither humans nor machines possess a definitive answer.
Do not search for what AI cannot do.
Calculation was uniquely human until it was not. Chess was uniquely human until it was not. Language, creativity and reasoning followed.
Education cannot be rebuilt around a shrinking catalogue of supposedly unique human abilities.
What must humans understand, value and take responsibility for—even when AI can do it better?
A machine may propose a decision, but human beings must live with its consequences. It may generate an argument, but someone must decide what to believe. It may design a life, but it cannot inhabit that life.