What Feels Deterministic In Pixels Depends On Hidden Ordering Rules

I assumed first that order was of little importance in $PIXEL so long as it contained the same elements. Where two sequences have the same steps, the intuition is that they have a kind of convergence property. As soon as I would alter the form, without altering the substance, that presumption shattered.

I kept everything the same actions, same amount of total input I just varied the internal sequence. The result shifted in a non-random way. Some combinations consistently gave smoother continuity, and other combinations definitely led to minor imperfections which wouldn't be immediately apparent. Nothing failed immediately, but the system carried those differences forward.

That would seem to imply that Pixels is not treating actions as discrete units. It is more like an ordered system with some dependencies on the different “steps”, even if those dependencies aren’t obvious They clearly exist. Previous actions don’t simply output later actions they create the environment over which later actions are read.

This is more like a dependency graph than a simple loop. Since order determines context, the same things can mean different things depending on how they’re ordered. This isn’t just information processing it’s information inputing.

Stacked layered on top, this becomes more meaningful with this structure. In this case, behavior can be analyzed across many users, and ordering regularities can be numerically computed, compared, and indirectly optimized en masse without ever explicitly revealing those rules. To the extent that this path leads to the best solution over the course of the collaboration. Realizing this, it ceased making sense to think in terms of “what to do”.
$PIXEL #pixel @Pixels