Predict-then-verify

Say what will happen before you run, compile, or check it, so the outcome teaches you instead of merely surprising you.

When to use this: While you practise
Learning impact: Strong change
Learning evidence: Credible with caveats
Where you are with the topic: Building understanding
How often to use it: Worth repeating often

When to use it

You are about to run code, a simulation, or an experiment and usually just wait to see.
Worked examples make sense only after you peek at the result.
You want a stake in the answer before feedback arrives.
Lab write-ups punish “I observed…” with no prior expectation.

What to ask AI

Type something like: “Before I run this, I predict: [prediction]. Ask me why. Do not reveal the outcome until I commit. Then help me compare prediction vs result.”

Ask it to refuse the answer key until a written prediction exists.

After the result, ask what your prediction assumed that was wrong or right.

How it helps

A prediction turns a result into feedback on your model of the world.

Without a prediction, every outcome is just news. With one, every outcome is a test.

The learning event is the comparison, not watching the demo alone.

AI can hold you to the commitment and interrogate the mismatch.

How to run it

State the setup in one or two lines.
Write a concrete prediction (what you will see, return, or measure).
Add a one-line reason.
Only then run, compile, look up, or ask for the actual outcome.
Explain the match or mismatch before you rewrite the mental model.

Example exchange

You

I have a loop that appends i*i for i from 0 through 3. I predict the list [0, 1, 4, 9]. Do not confirm yet, ask me why I think that.

Why those four values, and why that length?

You

i runs from 0 through 3 inclusive, and each term is the square of i.

Run the code now. If it matches, good. If you had predicted [1, 4, 9], we would talk about off-by-one. Want a cousin snippet to predict next?

Copyable prompt

Setup: [code / experiment / problem].
I will write a prediction and a one-line reason FIRST. Do not reveal or hint the
outcome until I do. Then ask me to compare prediction vs result and name what my
prediction assumed.

The Tell

You will know the method has flipped when you “predict” after you already saw the output, or let the AI announce the result before you commit.

Then nothing was at stake, and the method was theatre.

Write the prediction in the chat first, freeze it, then run.

Principle evidence

Strength of the underlying learning idea, not a claim about AI products.

Generating an expectation before feedback is a form of generative / predictive learning. Comparing expectation to outcome supports updating your model more than passive observation.

AI delivery evidence

Whether an AI tutor delivers this method well is a separate question.

AI is useful as a commitment device and a mismatch interviewer. It is not required for the principle, a sealed note before you hit Run does the same job.