misomes.ai

Build enough to learn something.

A prototype is useful when it makes a question easier to answer. It does not need to resemble a finished product. Often, removing detail is what makes an experiment readable.

Choose one question

Instead of trying to prove that an entire idea works, isolate one decision. Does changing the spacing improve the shape? Can a single control explain the relationship between two values?

Keep a visible baseline

Save the simplest version of the experiment. Compare each change against it. Without a baseline, it is easy to confuse novelty with improvement.

Stop at an explanation

An experiment can be complete when you can explain what happened and why. The next step may be a new study, rather than more features in the same prototype.

See a small example

The parameter playground keeps its scope to two controls and one drawing.

Read more