Quizzes with distractors built from real misconceptions
A quiz is only as good as its wrong answers. A distractor nobody would pick teaches nothing; a distractor built from the misconception students actually hold tells you exactly what to re-teach.
Why the mainstream APIs fail here
Generated quizzes usually have three nonsense options and one right one, so the score measures elimination not knowledge. Building real distractors by hand is slow, and the misconceptions themselves are undocumented in most courses, so the quiz writer does not know what students actually believe.
How MagmaRouter is used for this
Ask for the question, then the misconception each distractor encodes, so you can see the diagnosis the quiz performs. Ask for the answer explanation to name the misconception. Generate at three difficulties and keep the ones that split a room.
A working loop
- Every distractor needs a misconception named; delete any that has none.
- Explanations should teach the misconception, not restate the answer.
- Keep a bank of questions that half the class misses; those are your re-teach list.
Prompts are not stored. A signed receipt of the hash comes back on every call, so you can prove later what MagmaRouter logged. Details are on the privacy page. Per-model prices are on /models and the pricing model is on /pricing.