Our approach
Good education never fights how children actually learn
We build on learning science and the world’s best project-based-learning practice — and we show our work: every principle below maps to a concrete mechanism you can inspect.
Learning science
Five laws of learning
These are mainstream, well-established principles of how children learn — and how each one is wired into our system.
- 01
Active construction
Children learn by doing and building, not by passively receiving. Understanding is constructed through real work on real problems.
In our system
The Project step of the daily four-step loop: every day, the child pushes a real project forward with their own hands — code, prototypes, drafts — not worksheets about someone else’s project. - 02
Retrieval & spacing
Knowledge sticks when it is actively retrieved and revisited over time — not when it is delivered once in a single lecture.
In our system
Foundation micro-lessons are dispatched against gaps the project exposes, then revisited in the weekly review — spaced, need-driven practice instead of one-shot cramming. - 03
Formative feedback
Feedback improves learning when it is specific, timely, and tied to the work itself — not a grade at the end.
In our system
Written mentor feedback every week with a ≤48h SLA, plus a monthly five-dimension report. Feedback follows the evidence: it always points at concrete artifacts. - 04
Zone of proximal development
Learning happens fastest on tasks just beyond what a child can do alone — reachable with support, neither trivial nor overwhelming.
In our system
We diagnose a baseline first, then start the path with stretch-but-reachable tasks. You can preview how the path adapts in the three-slider path planner. - 05
Autonomy · mastery · purpose
Durable motivation comes from choosing your own direction, feeling yourself get better, and doing work that matters to someone.
In our system
The child picks the project direction (autonomy), lights up skill gates on the map (mastery), and ships work that real users actually use (purpose).
Industry benchmarks
Benchmarked against the field’s best
We adopt the standards, we don’t borrow the brands: these organizations don’t endorse us — we hold ourselves to their published bar.
-
PBLWorks — Gold Standard PBL
Seven essential design elements for project-based learning, from a challenging question to a public product.
Our implementation
Our project template makes the seven elements required fields — a project cannot launch without answering each one. -
High Tech High — exhibition of learning
Student work is presented publicly, to a real audience, as a first-class part of the learning.
Our implementation
Demo Day: every cycle ends with the child presenting their work publicly to family and real users. -
IB CAS / Extended Essay — reflection portfolio
Structured, ongoing reflection documented alongside the work itself.
Our implementation
The daily Reflection step plus a full process archive — every version, every decision, kept. -
Mastery Transcript Consortium
Credentials as mastery claims, each backed by linked evidence rather than a single grade.
Our implementation
Every skill on the skill map links to the concrete evidence behind it — work, not assertions.
The consequences
What this makes us refuse
A method you actually believe in forces you to say no. Each refusal below follows directly from the laws above.
-
We refuse drill volume as a stand-in for understanding
Retrieval and spacing say practice must be need-driven and distributed. Piling on repetitive problem sets fakes progress; we dispatch practice against real gaps instead.
-
We refuse to judge only the final artifact
If learning is constructed, the process is the evidence. We keep and review intermediate versions — a polished result with no visible process proves nothing.
-
We refuse to promise outcomes
We never promise a school place or a test score. Learning laws describe how children grow, not a guaranteed endpoint — anyone selling certainty is selling against the science.
-
We refuse to let AI do the work for the child
If the child does not construct it, the child does not learn it. AI use is logged and distinguishable in the archive, so you can always see who did what.
See it running