Approach

Agency. Taste. Craft.

When AI can make almost anything, students need to know what's worth making, whether it's good, and how to finish it. Students practice all three on projects they choose.

Three pillars

How each pillar is practiced.

Agency

Choosing what's worth solving.

AI can help make almost anything, so the real work is deciding what's worth making. Students choose their own project on day one.

  • Pitch your own idea in three minutes, no slides.
  • Cold-email an expert and ask for fifteen minutes.
  • Sign a promise to self to keep for ninety days.
Principles I · V · VI
  1. IAgency from day one. Mentors support; students choose.
  2. VReflection, started early. Thirty screen-free minutes a day on one question: which of my goals are my own?
  3. VIRisk early, while it's cheap. Small, repeated risks make larger ones easier.

Taste

Telling which version actually works.

AI helps students try many versions. They learn which one works by showing their work to users and practitioners every day.

  • Put a prototype online at a public link by midway.
  • Pair-code with a YC-backed founder for two hours.
  • Test the project with five questions, then defend each answer.
Principles II · III · VIII
  1. IIMake things in front of people. Every day ends with a demo.
  2. IIIExperienced mentors in the room. Mentors are founders, engineers and designers who have made real products.
  3. VIIIFirst-principles thinking. Students are assessed on asking how, why, what principle, when it fails and how to adapt.

Craft

Finishing it well.

AI helps students reach a first draft sooner; finishing a reliable product is their own work. Every build has a fixed freeze and ends in a live demo.

  • Build all week, using AI coding tools and making every call.
  • Stop at the freeze and demo what works.
  • Write a one-page personal routine to take home.
Principles IV · VII
  1. IVAI as a partner students direct. Students write the prompts, question the output and check it for mistakes.
  2. VIIDaily habits, personal systems. A three-line update every day: finished, learned, next.
A typical day

One day in a cohort, start to finish.

  1. Start
    Lesson

    taught right before it's needed

  2. Most of the day
    Build

    a problem the team chose

  3. During the build
    Mentor 1:1s

    help while stuck

  4. Fixed hour
    The freeze

    no extensions

  5. Daily
    Reflection

    30 minutes, no screens

  6. End of day
    Standup

    finished, learned, next

  7. Last day
    Demo day

    a live pitch

  • Formats
  • Five-day intensives in school breaks
  • Weekends and after school
  • Partner cohorts at host schools
What students leave with
  • A working product with their name on it
  • A personal routine they wrote themselves
  • Professionals who know them by name
  • Experience presenting in public

Programs have run in Hong Kong in 2025 and Beijing in 2026. See the results →

Talk to us → How we work with schools