The small experiment comes first
An idea for keeping the first test small enough to teach you something.
Sample text for the reading-page design. This is not a published article by Anh.
A new idea tends to arrive with a whole product attached. There could be a dashboard, a history view, settings, and a polished onboarding flow. Yet the question underneath might be much smaller. Can this approach handle one real example?
That question deserves a test before the surrounding interface becomes a project of its own. The first experiment only needs to produce enough evidence to choose the next step.
Choose the uncertain part
Pick the assumption that would change the plan if it proved false. Maybe the input is harder to interpret than expected. Maybe the result needs more checking than the workflow can support. A test of the easy part will not answer either question.
Write down the input, the result you expect, and what would count as a failure. A short note is enough. It helps separate what the experiment actually showed from what you hoped it would show.
Let the result change the plan
A failed example can make the project more specific. It may reveal a useful boundary or suggest that a smaller feature is the better place to begin. The experiment has done its job if the next decision is clearer.
Keep the result beside the original question. Later, that pair will be easier to learn from than a polished explanation written after the outcome was already known.