Education is optimized for institutions, not learning. This is a manifesto for what works better.
1. Classes of 15, maximum
When everyone participates and learns from each other's process, you create conditions for actual learning. Large auditoriums were built for cost efficiency, not outcomes.
2. Collaborative and generative learning
Students, teachers, and AI form a single group. Everyone explores, generates ideas, and learns together, because no one has figured out yet how education should work in an AI-native world.
3. AI participates as a member of the group
All students use the same AI system that remembers the full class context: every problem tackled, every solution found, every process documented. The machine pushes humans to think beyond their first prompt. Humans push the machine beyond generic outputs. After three to five years, the system understands each student's thinking patterns, strengths, and gaps, then suggests career paths that match their actual profile.
4. Study cross-disciplinary foundations
Even in specialized programs, students learn what Devon Eriksen calls liberatory arts: logic, statistics, rhetoric, research methods, practical psychology, personal economics, and agency. Skills that compound across every domain.
5. Problems replace lectures
The teacher chooses which problems to tackle and in what order, starting simple, building toward chains that connect multiple disciplines. Students learn theory only when they need it to solve something real. This builds problem-seeking ability: knowing which problems are worth solving.
6. The teacher as curator
The role shifts from transmitting knowledge to curating problems and guiding the resolution process.
7. Tools without limits
If it helps solve the problem better, use it. Schools should pay for licenses or provide open-source alternatives.
8. Collective critique of solutions
When fifteen people solve the same problem with the same tools but get different results, the teacher highlights what works and what doesn't. This develops judgment: the ability to recognize quality in context.
9. Deconstruct winning processes
Students who produced the best solutions walk through every significant step: which tools they chose and why, where they got stuck, what decisions shaped the outcome. This becomes shared knowledge for the class and training data for the AI.
10. Projects replace theses
The capstone should solve a real problem in a new way, built with AI as cognitive co-pilot and the teacher as strategic guide. Not 200 pages no one will read.
This works with technology that exists today. Find people who want to learn differently, find a space, pick a real problem. Use every available tool, critique solutions collectively, document the process, repeat with harder problems. Education changes wherever people start doing this.