Software, not a tutoring service. There is no timetable, no travel and no fixed
pace. A child practises when it suits them, one question at a time, and every
answer comes back with the reasoning attached rather than a tick.
It covers the four things the paper asks for: reading inference, multi-step
quantitative reasoning, visual pattern reasoning, and written expression.
Every question ships from a reviewed bank of
3,900+
that travels with the app — not improvised while your child waits. Each one
passes a chain of automated checks before it can be served, and anything that
fails is dropped rather than softened.
Visual pattern questions get a second pass: the figure is checked for coherence,
and then a separate model is asked to solve the puzzle from scratch. If its
answer disagrees with the marked one, the question is thrown away. That check
exists because marking a wrong answer correct is the failure that would matter
most, and it is the one a language model is most likely to make.
The full method — the programme, the teaching approach and every check — is set out
in our approach.