What a signal should change: the place, the time, or the rules

The earlier articles tested only one way a signal can act - it tells the cell where the food is - and it paid only when it predicted the food early enough, and then only weakly (+0.012). This article asks which way pays: the place (where the food is), the time (how much food, everywhere), or the rules (what the cell itself pays). Every arm adds the same total resource per step, so the arms differ only in how the signal distributes or modulates it, and each arm runs on twenty seeds. The signal that changes the rules is four times stronger than the signal that says where (+0.0512 against +0.0117, with 18 of 20 seeds positive against 14 of 20); the signal that says when is indistinguishable from noise (+0.0080, 11 of 20). The reason is structural: the cells are immobile, planted in the poor centre of the grid, so a map is worth nothing to them, while a change in their own terms applies wherever they stand. Every printed number is re-derived from the deposited artifact by a machine check (20 checks, 0 failures).