Have the model quiz you on a topic
The model asks one question, waits, grades your answer honestly, then aims the next question at whatever you keep getting wrong.
Last updated
Fill in the blanks
Assembled in your browser — nothing you type is stored or sent anywhere.
Your prompt
Model-agnostic — works in Claude, ChatGPT, Gemini or any AI chat.
Why this prompt works
- →"One question at a time, then wait" is the instruction that actually makes this a quiz, instead of a wall of questions and answers the model generates by itself.
- →Directing follow-up questions toward whatever I got wrong turns the session into targeted practice instead of a fixed list unrelated to my actual gaps.
- →Honest grading has to be stated explicitly, because left unprompted, models default to waving near-misses through as correct rather than marking them wrong.
Part by part
- Act as the quizzer, not the explainer
- States the role, because the default behaviour is to teach. A model that explains while quizzing removes the retrieval effort that makes quizzing work.
- ONE question at a time, then stop and wait
- The instruction most often ignored without emphasis. A batch of ten questions turns active recall into a reading exercise.
- grade it honestly
- Counteracts the tendency to accept a near-miss as correct. An answer graded generously is one you will not revisit.
- weight later questions toward those
- Adaptive difficulty. Even coverage spends most of the session on material you already know, which is the least useful place to spend it.
- say so instead of guessing at an answer
- The guard that matters in a learning context, where a confidently wrong correction is memorised as fact.
Common mistakes
Quiz me on this topic.Produces ten questions at once and often the answers underneath. Reading a question and its answer feels productive and generates none of the retrieval effort that makes testing work.
Test my knowledge and explain the answers.Combining the two means the explanation arrives before you have tried to recall. The struggle before the answer is the part that produces learning, and pre-empting it removes the benefit.
Give me a multiple choice quiz.Multiple choice tests recognition, which is much easier and transfers poorly. Open questions are harder and are what you actually need when nothing prompts you.
Questions
Why one question at a time?
Because the retrieval attempt is the mechanism. Seeing question two while thinking about question one splits attention, and seeing all ten turns the exercise into reading comprehension.
Can it grade reliably?
For factual answers, generally yes. For nuanced or partially correct answers it errs generous, so treat a partially correct grade as a prompt to check the source rather than as a pass.
Why weight toward weak areas?
Because time spent on material you already know produces almost no learning. Concentrating on gaps is the same principle spaced repetition applies to scheduling, done within a single session.
Is this better than flashcards?
It is better at open-ended and applied questions, where a card cannot capture the range of correct answers. Flashcards remain better for atomic facts, because the scheduling is what makes those stick.