A flatworm cut into pieces regrows a complete worm from each piece. What decides how many heads it grows is not the genome but a bioelectric pattern the tissue holds, a memory of the target body. Change that pattern and the same genome builds a different worm, and keeps building it every time you cut.
Drag across a worm to cut it. Pieces smaller than a crumb die; everything else regrows.
Each step animates on the dish. Do them in order, or improvise with the instruments.
Every worm on this dish has the same DNA. The number of heads is set by a voltage pattern the cells hold together through gap junctions. Levin's lab made two-headed worms that keep regrowing two heads across rounds of cutting, with no genetic change.
Paint a small spot and it fades back: the neighbors pull it into line. Paint a broad region and it holds. A pattern memory belongs to the collective, not to any single cell.
The body here is an agent pursuing a goal in anatomical space. You change its behavior by rewriting the goal it is pursuing, not by micromanaging cells. That is Levin's case for working with goals instead of mechanisms in any intelligent system, including the AI kind.