Education for a world of abundant intelligence
When intelligence is abundant, the bottleneck stops being knowledge and becomes judgment, and judgment can be deliberately built.
Information is abundant. Expertise is abundant. Educational content is abundant. What remains scarce, and contested, is the framework you use to interpret any of it.
The same fact means different things depending on which model you are running: which incentives you prioritise, which trade-offs you accept, which uncertainties you tolerate. When information and intelligence are everywhere, the bottleneck is no longer knowledge acquisition. It is judgment.
The three responses
Faced with abundant information and competing frameworks, people tend to do one of three things.
The first is to reject expertise altogether: if everything is contested, then nothing can be trusted. The second is to outsource judgment, leaning on experts, institutions, influencers, tribes, algorithms or AI to decide what to think. The third, and the only durable one, is to develop the capability to direct intelligence: to run the right model when it matters, reason under uncertainty, decide when no answer key exists, and act despite incomplete information.
The case worth making is that the third response can be taught. Not as content, but as capability.
Judgment, not subjects
Traditional education organises itself around subjects: mathematics, history, science, literature. These matter. They supply knowledge and context. But they are not the objective. The objective is judgment, which is narrower and harder: the ability to identify what matters and ignore what does not, recognise patterns across domains, understand incentives and trade-offs, decide under uncertainty, learn continuously, and act effectively in the real world.
Knowledge is an input into judgment, not the destination.
The hidden curriculum, made explicit
Most educational systems carry a hidden curriculum. Alongside the content, they teach habits, assumptions and ways of seeing. The move worth making is to take that hidden curriculum and make it the point.
Across every subject, book and problem, the same enduring concepts recur: compounding and incentives, trade-offs and feedback loops, bottlenecks and adaptation, probabilistic thinking and systems thinking, leverage and second-order effects. Taught as isolated abstractions, they do not stick. Met again and again through physics, literature, history, personal finance and ordinary life, they become lenses. The subject changes; the underlying pattern holds. Over time, the patterns are how reality gets read.
The tutorial model
The innovation here is not subject matter. It is method. The model draws less on schooling than on the older tutorial tradition, which begins not with content but with a learner, and whose first task is to understand how that learner thinks.
The loop runs: problem, response, analysis, remediation, generalisation. The learner attempts a problem. Their response reveals what they know, what they assume, how they reason, and where they are wrong. The system then teaches what is actually needed, and generalises the lesson beyond the specific case. The answer becomes the curriculum; the conversation becomes the learning.
Interests as the way in
Most personalisation chases engagement. Interests can do something better than motivate: they activate the mental models a learner already holds and can extend. Someone drawn to football can meet forces through football; someone drawn to fantasy, through dragons; someone drawn to history, through medieval warfare. The context changes; the capability does not.
But the context is load-bearing, not decoration. You understand forces better through football because football activates prior knowledge of velocity, mass and impact. The teaching happens through the activation, not through sugar-coating.
Making thinking visible
Every learner builds a distinctive map of how ideas connect: what they believe, how those beliefs hang together, which are held loosely and which tightly, and how the whole thing changes over time. Make that map visible and it stops being a ranking and becomes a diagnosis, a way to see the architecture of your own thinking.
You cannot improve a mental model you cannot see.
The same visibility, applied to how a person interprets reality and navigates decisions, becomes a kind of fingerprint: the patterns they notice, the evidence they trust, the trade-offs they prioritise, the uncertainty they tolerate. Its use is not to prescribe beliefs but to catch you leaning on untested assumptions, avoiding evidence that contradicts a preferred model, or mistaking confidence for accuracy.
Judgment as an emergent property
None of this is a cloning system. The purpose is not to produce people who think alike. Judgment cannot be downloaded or memorised; it emerges from a loop. Accurate models of how the world works meet evidence that challenges them. Reflection exposes the gap between prediction and reality. The model is revised, and feedback establishes whether the revision was an improvement.
The goal is not that learners arrive at predetermined conclusions, but that they become steadily better at forming, testing, revising and acting on their own. Disagreement, on this view, is not failure. It is evidence of independent thought.
Why AI changes everything
Traditional systems have to encode the intelligence into the curriculum: content written, questions authored, assessments built, pathways specified. That is expensive, rigid, and slow to adapt.
AI moves the intelligence out of the curriculum and into the tutor. The curriculum object can then be simple, a topic, a set of objectives, some optional source material, while the tutor generates the explanations, discussions, analogies, cases, remediation, assessment and reflection. The curriculum becomes a declaration of intent rather than a detailed script. Tutorial education, one-to-one and adaptive and focused on the learner's reasoning, stops being boutique and becomes available at scale. That is why this is possible now, and was not before.
The industrial age rewarded obedience. The information age rewarded access. The age of AI rewards the ability to direct intelligence: to ask better questions, to see what matters, to act when no answer key exists.
The goal is not to know more. The goal is to see more clearly, and to act with conviction in the spaces where clarity does not yet exist.
In practice
Asymptote is an operationalisation of this essay: an adaptive learning engine built to develop judgment rather than deliver content. The argument above, turned into a working system.