← Back to selected work
LearningPrototype

H-AI Pass

Korean history quizzes that connect to wrong-answer records and targeted practice.

Implementation flow

  1. 01Quiz
  2. 02Result
  3. 03Wrong answers
  4. 04Practice

Diagram of the implemented structure, not a live run or a performance result.

The question

How can a wrong answer become a specific next study action?

What I built

Quiz, result, wrong-answer, and practice interfaces, with account/data code and server-side question-generation providers.

A technical decision

Keep generation providers separate from the quiz and review flow, so generation is one part of learning rather than the entire product.

AI’s role

Server-side provider code can request question generation. The accuracy of generated questions has not been established.

Current status

Prototype. Question accuracy, duplicates, access boundaries, and the current deployed end-to-end flow need further validation.