note
An experiment can disappoint you
A working definition for separating experiments from demonstrations with the answer already baked in.
An experiment earns the name when it can disappoint you.
That does not mean every small prototype needs a formal hypothesis and a clipboard. It means the act of building should expose some belief to friction. The interface may not teach what I thought it would. The data may refuse the clean story. The fast method may be slower. The elegant abstraction may make the ordinary case harder to understand.
If the result cannot alter the next move, it may still be craft, illustration, or play. Those are worthwhile. They are simply doing different work.