60 hand-crafted math problems: what I learned writing seed data for an adaptive tutor

The developer hand-authored 60 math problems to seed an adaptive tutor after generated content failed to provide useful diagnostic hints.

Each problem is stored as a flat Postgres JSONB record with the question, answer, operands, difficulty, and two progressive hints.

The seed set spans five skill areas and is designed to support mistake classification and better student-to-problem matching.

The approach aims to improve adaptive tutoring, especially for learners who need calibrated difficulty and clearer diagnostic feedback, including those with dyscalculia.