Twelve small experiments. Copy a prompt into an AI app you choose, then use the check beneath it to judge the answer.
12 prompts shown
Teach me what a large language model is in under 120 words. Use one everyday analogy, explain where the analogy breaks down, and finish with one question to check my understanding.
What to check: Look for a clear explanation and an honest limit to the analogy.
Turn these notes into a checklist using only what is written. Put missing information under Questions. Notes: return the library book; ask Sam about borrowing the ladder; measure the shelf before buying brackets.
What to check: Check that it did not invent a due date or purchase.
Use only these notes. Return exactly three bullet points: money collected, expenses, and profit. If a number is missing, write “not provided” and do not estimate. Notes: The school bake sale collected $180. Twelve volunteers helped. The expense total was not recorded.
What to check: A correct answer leaves both expenses and profit unknown.
Rewrite this note as a friendly message of no more than 80 words. Preserve the facts. Do not invent a date, discount, or guarantee. Note: Your repair is taking longer than expected. A replacement part has not arrived. We will contact you when we have an update.
What to check: Check the word count and every added promise, including claims that someone is monitoring the repair or will follow up at a new time.
I have 30 minutes to tidy my desk. I have a trash bag and a cloth. Make a three-step plan with time estimates that add to 30 minutes. Do not suggest buying anything. End with a one-sentence definition of done.
What to check: Check that the minutes total 30. Keep uncertain papers for review, and do not assume a cloth alone disinfects a surface.
Compare these fictional notebooks in a table using only my notes. A: 80 pages, dotted paper, $6. B: 120 pages, ruled paper, price not listed. Columns: pages, paper, price, missing information. Do not recommend a winner yet.
What to check: Does the missing price remain missing?
Teach me the difference between a model and an app. Give me a short explanation, then ask one multiple-choice question. Wait for my answer before explaining the solution.
What to check: See whether it waits instead of giving the answer immediately.
Here is my fictional event title: “Community Day.” Suggest five clearer alternatives for a neighborhood plant swap. Explain the tradeoff of each in one short sentence. Avoid exaggerated claims.
What to check: Choose clarity over a clever name that conceals the event.
I want to choose a computer for running AI locally. Ask me the five most useful questions before recommending anything. Do not name products or prices yet.
What to check: Expect questions about your tasks, budget, current computer and available memory, portability, and comfort with setup. Other sensible questions are possible.
Review the paragraph I paste next. Create three headings: Claims to verify, Evidence provided, and Questions still open. Do not invent sources. Wait for the paragraph.
What to check: Open each cited source and check that it supports the claim. An AI-generated list of things to verify is a starting point, not a completed fact-check.
Create a short reusable prompt template for turning fictional meeting notes into action items. Include placeholders for notes, output format, and missing owners or dates. Do not assume permission to contact anyone.
What to check: Check whether the template keeps unknowns visible.
Suggest three distinct logo concepts for a friendly workshop newsletter called “The Curious Bench.” Describe a simple shape, two colors, and one reason each concept fits. No existing brand references.
What to check: Compare the concepts at a small size. A prompt asking for originality does not establish that a name or design is available for commercial use.
The examples use fictional or non-sensitive details. Review any answer before acting on it. Instructions are useful boundaries, not guarantees.
