Sticking together is a strategy.Predation is only one reason to do it.

What the documents got right, and the one place they tunnel-visioned.

The substrate was never the problem. §1 carries an adhesion and a sharing effector and conserves matter in transit; §3 lists kin-gated adhesion → multicellularity and kin resource-sharing. The pieces for several distinct benefits are already in the design.

The justification is where it narrowed. Every place the docs say why colonies should pay, they cite predation only — the size-refuge, the engulf gate as the "Boraas lever." We never wrote down the broader driver taxonomy, so the build inherited "predation-driven multicellularity" as the north star and the validation was constructed around that single lever.

Root cause: Boraas / Herron / Becks are the iconic, cleanly-validated origin-of-multicellularity experiments, so "ground it in real biology" pointed straight at predation. We ran a proper driver survey for the trophic/guild side (trait-based plankton models) — but never the equivalent survey for grouping.

Tong, Bozdag & Ratcliff sort the drivers of simple multicellularity into three functional families. Each is listed below with the EvoLab mechanism that would express it — and whether that mechanism exists yet.

Function 1Self-protection

• Predation size-refuge. A bonded unit is too large to engulf. In predator experiments, single-celled algae shift to stable small colonies that are effectively invulnerable within tens of generations. → engulf gate reads colony-effective mass · IMPLEMENTED
• Stress shielding. Outer cells buffer the interior from external stress (sharp gradients, toxins, light extremes). → emergent from the field gradients a cluster sits in · NOT YET TESTED

Function 2Resource acquisition & production

• Cooperative use / sharing. Cells that release or exchange resources do better clustered — a neighbour captures what would diffuse away, and reserves can move between bonded cells. → sharing effector (§1/§3) · DESIGNED, DEFERRED IN CODE
• Scarcity advantage. When the limiting resource is scarce, a group that pools energy beats single cells whose fixed upkeep eats their whole budget; when it is abundant, the single cell wins. A built-in environmental switch for when grouping pays. → falls out of shared reserve + DEB upkeep · NEEDS SHARING
• Niche escape (indirect). A group can occupy or exploit space the unicellular ancestor competes poorly in — grouping can be favoured even with no direct per-cell gain. → emergent from spatial competition · OBSERVE IN LONG RUNS

Function 3Dispersal & positioning

• Holding position. A lone cell in a good spot drifts out of it; a cluster holds station or anchors. This is a real benefit — not the artefact I first called it — and our sim already shows it. → bond springs damp the random walk · INCIDENTAL TODAY, MAKE IT HONEST
• Sinking / buoyancy control. Aggregate size changes sinking rate — a cost or a benefit depending on where the resources are. → couples to buoyancy effector + POM sink · PARTIAL

Grouping must not be free, or it is a bonus, not a strategy. These standing costs are what let a solitary cell win when no benefit above applies — so the choice to group is a genuine tradeoff the environment decides.

• Surface-to-volume. Packed cells share boundary layer and have less free surface per cell → slower per-cell uptake. The canonical cost of being interior.
• Self-shading. Cells in a cluster shade one another from light.
• Local depletion. A cluster draws its shared neighbourhood down faster than dispersed cells do.
• Bond upkeep + transport. Maintaining bonds and moving shared matter costs energy (already in §1, conservative).

The current crowding cost (cells resting at touching distance) bundles S/V, shading and depletion together and ties them to physical spacing — which is also what sets colony size and protection. These need separating so cost can be tuned without changing how large or how protected a colony is.

The single table the redesign works from. "Status" is what the model can express today.

How this reshapes the build, kept to the Charter: enable the structure, let evolution pick the driver.

• Grouping is one evolvable strategy with many payoffs. The model's job is to make each benefit expressible and each cost real; which driver carries multicellularity is the environment's call, run by run.
• Bring sharing and positioning to first-class alongside protection — not "later stages." A test with only the protection lever can't tell us anything clean, as Stage D showed.
• Separate the costs from spacing. S/V, shading and depletion should be tunable independently of colony size and protection.
• Boraas becomes one validation, not the target. Predation refuge, scarcity-sharing, and positioning each get their own pass/fail check; success is "grouping emerges under the driver its environment favours," and predation is the cleanest of several.