Four conditions for a self-maintaining organisation in a digital medium

What does a collective of digital cells need to persist, reproduce, and become subject to selection? We test four conditions in a world of cells that pay energy to exist: a reproduction unit smaller than the whole, a price for a heritable trait, a rule for when two types meet, and inheritance of the environment. Each is a pre-registered criterion compared with a null model, not with zero. Results. Selection was measurable only with single-cell reproduction: the opportunity for selection is 0.5614 against a null 95th percentile of 0.0621, exceeding it in 10 of 10 seeds, while the pre-registered lineage gate is met in 4 of 10. With colony-level or pair-level budding only one lineage holds data, so the opportunity is zero by construction and those worlds are called not evaluable, not evidence against selection. A pair of any two live cells does not help. The price of a trait is measurable only where lineages are numerous: median partial rank correlation -0.478 (8 of 10 valid runs). Collision is overwriting, not competition. Environmental inheritance does not exist: the advantage of a dying cell's neighbourhood is explained by the number of living neighbours. Conclusion. An organisation here is held together by geometry and by the speed of local updating, not by memory or by superiority; the unit of reproduction sets how many lineages coexist, and therefore whether selection can exist at all. This is a preprint; the paper is not peer reviewed. LaTeX source and the four figures are included. Code and reports: doi:10.5281/zenodo.23648667; checkpoints: doi:10.5281/zenodo.23654174. v1.1 (27 Oct 2026): new section 4.6 on transfer beyond bodies (R1 world without a grid: «one cell beats many» confirmed 10/10; monotonicity in unit size and the effect size do not transfer; the growth-limit explanation was tested and removed), plus updated limitations (price as a design constraint, environmental inheritance closed by a mechanism).