About CREATE BASE
CREATE BASE abstracts problem-solving operations and builds reusable systems from them. It started in education: an expert who sees the solution the moment they look at a problem broke their own judgment down into what to notice and what to do, and organised it into a form that works for other problems.
What we do
Extract the operations used to solve a problem, abstract them by removing the numbers and circumstances specific to the problem, and organise them against existing operations. Turn the organised operations into data, and generate explanations, flowcharts, diagnostics and other outputs from it. Running this system across domains is the core of our work.
- Make operations describablePairs of cue and response, and the connections between operations. The flow of judgment for each problem, recorded in a form anyone can read.
- Turn it into data, and generateFrom the data of organised operations, generate explanations, figures, animations, diagnostics and exercises in the same form.
- Cross domainsCue and response are not written in the vocabulary of any subject. The system confirmed in arithmetic extends to physics, law and every other domain.
- For people and for AIUsed for human learning; a prototype that extracts operations from AI-generated solutions is under way.
Why we can
The more expert someone is, the less visible their own judgment is to them: a person who sees the solution instantly is not aware of the steps in between. That is why structuring judgment has been hard. At CREATE BASE, that expert broke their own instant judgment down into cues and responses and systematised it against a single standard. Because every operation is cut by the same standard, it can be used in the same form across domains. A patchwork of materials, or explanations produced by AI one problem at a time, does not give this consistency.
Publishing policy
- Only what has been checked is called confirmed. Outputs are published after review by an expert. Unreviewed items are marked as draft, candidate or under analysis.
- Facts and inferences are kept apart. We do not add conditions or facts absent from the source. Learner remarks and AI remarks, party claims and court findings, are distinguished.
- No unsupported claims of effect. What is proven and what is a possibility are stated separately. Use for AI is at the stage of prototype and possibility.
- Sources are cited. Third-party problems, figures and opinions are handled through our own summaries and redrawn figures, with the source stated.
Where the system was proven
The system was used and tested in preparing students for some of Japan's most selective junior-high entrance exams, taken at age 12 and designed to test reasoning rather than speed. What it built was not subject-specific technique but the ability to judge for oneself.
- Students began to learn on their ownStudents who had internalised the structure of operations did not wait to be taught; they found problems and learned ahead by themselves. One example: a 15-year-old who earned Cisco's CCNA certification entirely by self-study.
- Admissions to the most selective schoolsNada Junior High School (consistently among the top feeders to the University of Tokyo), Tōdaiji Gakuen, Rakunan, Osaka Seiko Gakuin and others.
- First place with full marks in every subject on a major cram school's mock exam; several medallists in the Japan Math Olympiad for juniors