Give the next experiment one thing worth learning.

An experiment becomes more useful when you can say what it is meant to reveal.

That might be whether a person understands a representation, whether a mechanism can do a particular job, or whether an unfamiliar combination has enough promise to develop further.

Start there. Then choose the smallest adequate examination of that question.

Write the question before choosing the artifact.

“Build a prototype” describes activity. “Find out whether someone can recognize the consequence of changing this assumption” gives the prototype a purpose.

A paper model may answer one question. An interactive sketch may answer another. A technical proof of concept may be necessary for a mechanism that cannot be judged from appearance.

The useful form follows what you want to learn.

Include an outcome that would surprise you.

What would make you reconsider the idea? What result would suggest a different direction? What could reveal that the experiment itself was asking the wrong question?

Keeping those possibilities open helps an attractive artifact become an instrument of inquiry.

Keep the conclusion proportionate.

If one person understood the sketch, you learned something about that encounter. If a mechanism worked in one condition, you established something about that condition.

The result can be valuable without settling the whole undertaking.

Record what happened, what you infer, and what you still want to examine. Those are three different parts of the creative record.

Read human and AI contribution, constraints, and the first sketch.

Let the work suggest the next move.

You may improve the idea, widen the test, ask another specialist, or set the branch aside for later. A useful experiment can make any of those choices more intelligible.

Howard Pearlmutter's practice connects exploration with implementation. See human interaction with simulation, the Telos engagement, and a guided encounter.

One clear thing learned can give an unfinished idea a much better next step.