Re: To paraphrase Arthur Dent:
They are clumps of cells and not cell fragments. The stem cells, if left to their own devices in a suspension, clump into spheroids of epidermis with the outer cells producing cilia by which they swim. As far as I can make out what they've done is work out Conway-like rules about how they associate in clumps and from that have worked out a shape - something along the lines of the glider gun - which will assemble smaller similar shaped clumps from loose cells. They then carve up the self-assembled spheroids into this shape. The reconfigured clumps work as predicted.
Each generation, however, is smaller. As they don't actually grow the limit is reached with a generation that's too small to assemble another one. They don't acquire energy from external sources so even the unmolested spheroids will eventually fall apart. The system is not self-sustaining.
It looks as if it depends on some mechanism for interaction between cells which must be related to the mechanisms for growth, embryogenesis, repair and differentiation in the whole organism but which lacks the structure of the whole organism to direct it. It may offer some insight into how collections of single cellular organisms could come together to form simple invertebrates with more advanced behaviours.