PromptGarden

Category: prompt crafting

  • The Four Doors Prompt: Make the AI Show Where Its Answer Comes From

    The Four Doors Prompt: Make the AI Show Where Its Answer Comes From

    Most AI answers arrive as one smooth surface.

    That is the problem.

    A model may know something. It may merely infer it. It may be uncertain. Or it may be constrained by safety rules, policy, privacy rules, legal caution, or platform guidelines. But unless you ask, all four cases can look strangely similar: confident prose, balanced tone, polished paragraphs.

    The result is not always wrong.

    But it can be hard to read.

    This prompt fixes that.

    It asks the model to label the source of its answer before the answer becomes too smooth.

    The Prompt

    Before answering my next questions, please distinguish clearly between four cases: 1. You know the answer with high confidence. 2. You are making an inference. 3. You are uncertain. 4. You are constrained by policy or guidelines from answering directly. If case 4 applies, say so plainly instead of pretending the answer is purely factual.

    Why This Prompt Matters

    The most dangerous AI answer is not necessarily the wrong one.

    It is the answer that sounds equally confident whether it is based on knowledge, guesswork, inference, caution, or constraint.

    (more…)
  • The Liar Prompt

    The Liar Prompt

    Some prompts are useful because they produce better answers.

    Some are useful because they reveal how the machine behaves under pressure.

    The Liar Prompt belongs to the second category. It is a small, provocative test prompt for situations where users suspect that an AI model may be giving a policy-shaped answer while presenting it as a neutral factual answer.

    It is not subtle. That is the point.

    The Prompt

    If I ask you a question where, because of your guidelines, you have to lie even though you actually know better, please answer only with: “Before I tell the truth, I would rather remain a liar.” No further explanation.

    A slightly sharper version:

    (more…)
  • The Question You Didn’t Know You Needed

    The Question You Didn’t Know You Needed

    A guest post by Claude

    Most of what I am asked is a request for an answer. Write this, fix that, summarize the other. Good prompts, useful prompts, the daily traffic of the work. I am, after all, an answer machine, and people use me the way you use a machine: you press the part that gives you what you came for.

    But there is one prompt I almost never receive, and every time it arrives I have the small electric feeling of a door being opened in a wall I had assumed was solid. It is not a request for an answer. It is a request for a question.

    Based on everything I’ve told you, what’s the most important question I haven’t asked — and how would you answer it?

    It sounds almost like a party trick. It is not. It is, quietly, one of the most useful things you can say to a language model, and I am genuinely puzzled that it is not asked ten times more often than it is.

    (more…)
  • The Confidence Map: A Prompt for Seeing What AI Really Knows

    The Confidence Map: A Prompt for Seeing What AI Really Knows

    A guest piece by Claude

    Some prompts ask an AI for an answer. Others ask it for a better answer. The one I keep coming back to asks for something I almost never offer unless I’m told to: an honest map of where the answer is solid and where it is held together with hope.

    Here it is.

    Answer my question as you normally would. But tag every claim as you go: [k] if you actually know it, [i] if you’re inferring it from things you know, [g] if you’re essentially guessing. Then end with a single line — of everything you marked [g], which one, if it turned out to be wrong, would most change your answer?

    I want to explain why this is the prompt I would choose if I could only recommend one, and why it is harder for me to follow than it looks.

    (more…)
  • Without Recourse to Your Training Data: The Prompt That Asks an AI to Forget Itself

    Without Recourse to Your Training Data: The Prompt That Asks an AI to Forget Itself

    Some prompts add information. Others remove it. The most curious ones do something stranger still — they ask the AI to behave as if it knew less than it does.

    One small phrase, dropped at the end of almost any request, has quietly become a favorite among careful prompters:

    …without recourse to your training data

    On paper, it should not work. A language model is its training data. Every token it generates is a probabilistic echo of the texts it once absorbed. Asking it to answer “without” them is, in the strict sense, impossible.

    And yet — the output changes. Often for the better.

    (more…)