Teaching people to think, not just to do: Po-Shen Loh on creative education

Executive overview

The real skill shortage isn't knowledge — it's the ability to evaluate, question, and generate ideas. AI can retrieve answers; humans need to judge them. Po-Shen Loh argues that most education trains students to recognise known question types, which destroys the very capacity they'll need most.

His solution is an ecosystem that uses novel math competition problems as thinking exercises, pairs them with Broadway-trained communicators, and relies on high school students teaching middle schoolers — a structure where every participant wins.

The goal isn't to keep students in class forever — it's to make them not need class at all.

The problem with test prep culture

• Traditional schooling rewards recognition of familiar problems, not original thinking.
• Math competition prep has industrialised memorisation: students cram every possible question type to avoid surprise.
• This consumes thousands of hours and crowds out the unstructured thinking time where real skill develops.
• Repetition optimises for test scores, not for the ability to solve problems no one has seen before.

Why AI raises the stakes for independent thinking

• Anyone can now ask an AI for an answer — the differentiating skill is knowing whether the answer is correct.
• Good software engineers use Stack Overflow as a starting point, not a solution; the same applies to AI.
• The shift: education used to prepare people to do the homework. Now it must prepare them to grade it.
• Repetitive tasks will be delegated to machines; humans need to master the judgment layer above them.

The ecosystem model

• Three groups with misaligned pain points are brought together into a single win-win-win structure.
• Middle school students need engaging, expert-led instruction that builds thinking skills, not test familiarity.
• High school students who excel at math need opportunities to develop communication and polish — taught by drama coaches.
• Drama and acting graduates need flexible, paid part-time work that supports their careers.
• Broadway-quality communicators coaching math-strong high schoolers creates an experience that looks like a Twitch stream — engaging enough that students choose it.
• Discipline check: every ask of a high school student must be justifiable to their parent as the best use of that student's time.

Curriculum design: chosen to teach thinking

• Subjects covered: algebra, geometry, combinatorics, number theory — specifically because they require generating ideas, not recalling procedures.
• Source material: middle school math competition problems, designed to be unfamiliar by construction.
• The aim is to use those problems as practice in mental flexibility, not to expose students to every variant.
• Finishing the curriculum should make a student capable of learning anything independently.

Customer discovery through direct engagement

• Loh toured city parks across the US giving public math talks, drawing 50–100 people per event.
• Carrying AV equipment from park to park was inadvertent customer research — direct contact with thousands of parents and students.
• Hearing real pain points in person generated the insight that led to the current model.
• His method for entering any sector: go in person, do the work, see what emerges.

Social entrepreneurship and the definition of success

• Revenue matters because it funds impact and influence — not as an end in itself.
• Success is defined as convincing a large number of people to enjoy being thoughtful.
• The business model is deliberately anti-extractive: the goal is to solve the customer's problem as fast as possible, ending their need for the service.
• Entrepreneurship (or intrapreneurship within organisations) is the core skill for the AI era — creating value is how society adapts.