Predict Before Running

Write down what you expect before executing a test, query, benchmark, or probe. A result only teaches when it can disagree with a model.

Test

The question

What do I expect to observe, and why?

Do this

Make the prediction specific enough to be wrong.
Include the mechanism that produces the expected result.
Investigate disagreement instead of editing the prediction after the fact.

Example

Prediction: two concurrent decrements guarded by stock > 0 will let one buyer succeed and reject the other. Observation: both succeed. The model or query is wrong.

Ask next

?What result would refute my model?
?Which mechanism produces my prediction?
?What changed when prediction and observation differed?

← Use Concrete Examples State What the Test Does Not Prove →