thinkhard.SI

Prompt repository Research & learning

Test yourself on your NotebookLM sources with questions that go past recall

You have been learning from sources in a NotebookLM notebook and want questions that show whether you understood them, with a key to check after.

Registry no.
RES-051
Collection
Research & learning
Laid out for
Gemini
Score
10 out of 10
Blanks to fill
5

Research & learningGemini

The prompt

Score 10 of 10, Excellent

Use it when You have been learning from sources in a NotebookLM notebook and want questions that show whether you understood them, with a key to check after.

  • {{topic}}
  • {{goal}}
  • {{sources_to_cover}}
  • {{my_context}}
  • {{question_count}}
Full prompt and why it works
Prompt
I have been reading the sources in this notebook to learn about {{topic}}, so that I can {{goal}}. Write questions that test whether I actually understood them, not only whether I remember the wording. I will answer on my own first and check the key afterward, so keep the questions and the answers apart.

Sources to cover: {{sources_to_cover}}
My work or situation, for the apply questions: {{my_context}}
Number of questions: {{question_count}}

Spread the questions across these types, labeled with the type, in roughly this mix:
- Recall (about a quarter): a key fact or definition the sources state.
- Explain (about a third): why something is true or how it works, answerable in two or three sentences of my own.
- Apply (about a quarter): a short scenario from my situation above, where I have to use an idea from the sources to decide or predict something.
- Connect (the rest): compare what two sources say, or spot what the sources leave unsettled.

Format:
Part 1, Questions: numbered, with no hints or answers, from easiest to hardest.
Part 2, Answer key, with the line "Stop here until you have answered" above it: for each question, a model answer in under 60 words, a citation to the source it comes from, and one line on what a weak answer would usually miss.

Make sure every question can be answered from these sources alone, and that each Apply scenario has one answer the sources can defend. Avoid trick wording and questions that give away their own answer. Leave out multiple choice, because writing my own answer is what shows understanding.

For example, an Explain question in the right form: "Why do the sources say the second approach costs less over time, even though it costs more to set up?" That example shows the form only.

When I paste my answers in a later message, mark each one Correct, Mostly, Partly or Missed against the sources, cite the passage I should reread, and do not rewrite my answer for me. If the sources are thin on a type of question, write fewer of that type and tell me why instead of going beyond the sources.

Variables

{{topic}}
What you are learning, for example "how our new expense policy works" or "basics of supply chain finance".
{{goal}}
What the learning is for, for example "explain it to my team on Monday" or "pass the internal certification".
{{sources_to_cover}}
All sources in the notebook, or the ones to use by name.
{{my_context}}
Your job or situation in a sentence or two, so the apply questions feel real.
{{question_count}}
How many questions, for example 12.

Why it works

  • Length. A set mix of recall, explain, apply and connect questions tests understanding, not only memory, and the count keeps it to one sitting.
  • Format. Questions and the answer key are kept apart, so you answer first and check after, which is what makes a self-test work.
  • What to avoid. When you send your answers, it marks them against the sources and points you to the passage to reread, without rewriting your answer for you.

How to use it

  1. Press Fill in blanks and type your details into each blank. What you type stays in this browser.
  2. Copy the filled prompt into the AI tool you already use, such as ChatGPT, Claude or Gemini, or open it in the Grader or the Prompt Tester here.
  3. Before you rely on it, try it on three or four real examples. The Prompt Tester runs them all at once and shows which ones go wrong.
  4. When the answer comes back, check it before you use it. The Fact Check List lists every figure, quote and source in the answer worth checking, riskiest first.
Use it from code

Get this prompt as JSON from the free REST API, with your own text in the blanks, or let an AI assistant fetch it through the MCP server. No key.

curl -s https://thinkhard.si/api/v1/prompts/notebooklm-test-what-you-learned

With the MCP server added to Claude, ChatGPT, Cursor or VS Code, call get_prompt with {"prompt": "notebooklm-test-what-you-learned"}, or just ask for it by name.