Isn't this just more work for me?

It front-loads work you were going to do anyway as rewrites. Four answers up front is cheaper than three rounds of “no, not like that” — and you understand your own task better at the end of it.

What if I do not know the answers to its questions?

Say so explicitly: “I don't know 2 and 3 — give me a version that works either way and flag where the two paths differ.” That is a far better outcome than a silent guess.

Prompt foundations

Ask me questions first

Instead of letting the model invent the details you forgot to mention, make it ask you for them before it writes a single line.

Use this when the task is big enough that one wrong assumption early on wastes the whole output.

The pattern

Copy this and replace everything in [SQUARE BRACKETS] with your own specifics.

Prompt template

I need [THE OUTCOME, one line].

Before you write anything, ask me the [3-5] questions whose answers would most change what you produce. Ask only questions you genuinely cannot infer from what I have told you. Number them, then stop and wait.

Once I answer, produce [THE DELIVERABLE] and tell me which of my answers changed your approach.

Why each part is there

Before you write anything: Without this the model helpfully asks its questions AND answers them itself with invented facts — worse than not asking, because the guesses are now buried inside a finished draft.
Most change what you produce: Ranks questions by leverage. Otherwise you get five polite clarifications about tone when the thing that actually matters is who the audience is.
Then stop and wait: An explicit halt. These models are tuned to be maximally helpful within one turn, so they need permission to hand the turn back to you.
Which answers changed your approach: Makes the reasoning inspectable. If none of your answers changed anything, the questions were theatre and the prompt is worth re-running.

Before and after

The lazy version

Help me plan my Class 11 study schedule for the next three months.

What goes wrong: The model does not know your board, your subjects, your weak chapters, your exam dates, or how many hours you actually have. So it invents all five — and produces a confident, beautiful schedule built for somebody else.

The pattern, filled in

Worked example

I need a three-month study schedule for Class 11 that I will actually follow.

Before you write anything, ask me the 4 questions whose answers would most change what you produce. Ask only questions you genuinely cannot infer from what I have told you. Number them, then stop and wait.

Once I answer, produce a week-by-week schedule and tell me which of my answers changed your approach.

What that buys you: You get asked about your board, your two weakest chapters, your real free hours on weekdays, and your exam dates — and the schedule that follows rests on four facts instead of four guesses.

Now do it on your own work

Pick a task where AI output has previously disappointed you. Run this pattern. Write the model's questions down BEFORE answering them, then honestly mark which ones you had left out of your original prompt.

Check yourself — every box should be true

The model asked questions and actually stopped, rather than answering itself.
At least two of its questions were things you had genuinely left out.
You can name one question that was useless, and say what the model could have inferred instead.
The final output reflects your answers — you can point at the line that changed.

Then explain it back

Which question surprised you, and what does the fact that you had not thought of it tell you about how you were framing the task?

Write two or three sentences in your own words. If you cannot, you copied a prompt rather than learning a pattern — which is the whole difference this library is trying to make.

Does this depend on which AI tool I use?

Reasoning-focused models usually respect the halt; chattier, heavily helpfulness-tuned ones often need it repeated. If a tool keeps barrelling ahead, move “Then stop and wait for my answers.” to the very last line of the prompt, closest to where generation begins.

Why it actually works

This is the same instinct as our tutor's hint ladder: an answer handed over before the problem is understood teaches nothing. Surfacing the gap is the whole game.

How our hints-not-answers ladder works →

Prompting well is a skill you can pick up in a week. Understanding what is happening underneath is the 18-month curriculum — and it is what stays useful when the tools change.

FAQ

Isn't this just more work for me?: It front-loads work you were going to do anyway as rewrites. Four answers up front is cheaper than three rounds of “no, not like that” — and you understand your own task better at the end of it.
What if I do not know the answers to its questions?: Say so explicitly: “I don't know 2 and 3 — give me a version that works either way and flag where the two paths differ.” That is a far better outcome than a silent guess.

Next patterns

The Prompt Patterns Library is a separate, optional track — it does not change the 78-week curriculum or your roadmap. See what the curriculum actually covers →