PAIMON / Method

Curriculum vs scraping

The curriculum is the product. The model is what happens when you teach it.

How a subject is broken down

Every subject becomes a folder of lessons, one markdown file per lesson, each written to stand on its own. Lessons are ordered so that nothing depends on something the learner has not been given yet. The folder is the syllabus and the syllabus is version controlled.

The teaching loop

Lessons are taught in batches. After a batch, the system is tested on held-out passages: material from the same subject that was deliberately kept out of the teaching set and marked so it can never be taught by accident. That exclusion is enforced in the pipeline, not left to discipline.

A batch that fails its test does not get quietly accepted. The lesson goes back for a rewrite and the held-out material stays held out. Lowering the test to make the numbers look better would defeat the entire exercise.

What the memory bank holds

What passes is consolidated into the memory bank: the lessons, the passages and the provenance that ties an answer back to the lesson it came from. Provenance is the whole point. An answer you cannot trace is an answer you cannot sell.

Scraping, by contrast

Scraping is cheap, fast and unauditable. You cannot say what is in the pile, you cannot remove one bad book, and you cannot prove the system was ever taught the thing you are charging for. We would rather spend the time writing the lesson.