Observations between experiments
Not every useful discovery needs a full experiment entry. These are the smaller things worth remembering.
Signal Garden looked most impressive when everything drifted. It became useful when familiar clusters stopped moving.
The change was simple: once a cluster had been visited twice, it earned a fixed position. The visualization lost some spectacle and gained a usable memory. That trade was correct.
A prototype gets easier to evaluate when the reason to pause, fail, or graduate is written before the build begins.
More input can make an AI answer sound complete while quietly making its evidence harder to inspect.
Weekend Window confused availability with willingness. The prompt solved the wrong problem and made the experience worse.
Abandoning the notification loop was not a failure to finish. It was the result the test produced.
The gesture is expressive, but the workload scales badly. The idea is waiting for a context small enough to deserve it.