In / outsensing & effectors

The channels the genome reads to drive its actions. Principle: hand cells raw physical quantities, never labels — the GRN decides what a reading means.

What a cell reads

• Internal: reserve C, reserve N, body size. (Reserve fullness is derivable.)
• External scalars (the value where it is, sampled over its footprint): light, inorganic C, inorganic N, and the marker channels.
• External gradients (which way each improves): light, C, N, markers — but the direction is size-limited (below).
• Short-range contact: is another cell within reach, and how big / rich is it — the input predation will read (detailed in 3.2).

No labels: a cell senses "nearby cell biomass," never "predator." No free clock: day/night arrives only through light; a circadian rhythm can emerge if the genome is recurrent (§5). Senses are free — the only cost is emitting a marker, which is an effector (3.2).

Direction sharpens with size

A cell always knows the value of a field where it sits. But knowing which way a field improves means comparing across its own body — and that comparison's signal-to-noise rises continuously and steeply with size. So a tiny cell gets a near-useless, noisy direction; a large cell gets a sharp one. No hard switch — a curve.

Two things fall out. Small cells, to navigate at all, must average the noisy reading over time as they move — which is exactly real run-and-tumble chemotaxis, and it emerges for free (needing only GRN memory, §5). And it stays non-steering: size — itself emergent from §1 — sets how much to trust direction; the genome still decides what to do.

Markers are a barcode — and that's how kin recognition emerges

Instead of one shared signal, the marker is a small set of ~3 diffusing channels — a continuous "tag-space." Each cell carries a heritable, mutable emission profile (its barcode) across those channels, so lineages drift apart in tag-space over generations. A cell senses the per-channel concentration and gradient nearby.

Because the genome co-encodes both the tag a cell emits and how its GRN responds to tags it senses, kin recognition, kin-gated adhesion (→ multicellularity), discriminate predation (eat the dissimilar), kin resource-sharing — and cheating — all emerge. This is the real "greenbeard" mechanism: a locus that carries a cue, recognises that cue in others, and acts on it. In Dictyostelium and Myxococcus it's what gates aggregation, resource sharing, and the transition to multicellular life. Nothing about kinship is coded — we supply only the barcode substrate.

What a cell can do — each action drawing on the §1 budget. Two rules hold it together: the guild emerges from how & what you eat (never a coded "predator"), and every effector pays a cost — nothing is a bonus.

Metabolic

• Uptake — absorb dissolved C/N from the local field, inorganic or organic (osmotrophy folded in). Cost: energy; rate scales with surface.
• Fixation — light + inorganic C → organic C into reserve. Gated by light; η_fix < 1.
• Respiration — organic C → inorganic C + energy. η_resp < 1 — the night / energy path.
• Excretion — return surplus C/N to inorganic. Automatic on overflow (the §1 hoarder fix); can also be GRN-active.

Movement

• Swim — set a direction + effort (any direction). Cost ∝ effort × size (drag); a stochastic reorientation turns small cells' noisy direction-sense into run-and-tumble.
• Buoyancy — cheap, vertical-only rise/sink. Slow and up/down only, but cheap — so diel vertical migration can evolve without paying full swimming cost.

Trophic — the guild lives here

• Engulf (phagotrophy) — ingest an adjacent particulate source (a live cell or a detritus particle). Feasibility = attacker size vs the target's colony-effective size (graded, strong cutoff): solitary cells are edible, a bonded colony is one big un-engulfable unit. η < 1; cost: energy.
• Secrete-enzyme (extracellular digestion) — convert adjacent particulate organics into dissolved organics in the field, which Uptake then absorbs. The bacterial/fungal decomposer route — and a public good: nearby osmotrophs can absorb the freed matter without paying, so cheating is possible.

Osmotrophy isn't a separate effector — it's Uptake acting on dissolved organics. So the three routes to organic carbon are engulf particulate, absorb dissolved, and secrete-enzyme to turn one into the other.

Reproduction

• Divide — split, conservatively partitioning structure + reserve + energy. The split can be asymmetric (GRN-set fraction), so differentiation and division of labour can emerge. Mutation happens here (§5).

Social / multicellular

• Adhere — form or break a spring-bond with a neighbour. Cost: bond upkeep (structure). Whether to bond, and to kin, is GRN-decided off the barcode sense.
• Transfer — move reserve across a bond to a partner. Matter conserved in transit; cost: per-unit transport energy.
• Emit marker — secrete the heritable barcode into the channels. Cost: energy; rate is GRN-modulated.

The hook into 3.3: a cell can't run all of these at full tilt — the machinery for each draws on one finite capacity. 3.2 names the actions and their costs; 3.3 makes the split zero-sum, which is what forces specialists to beat generalists.

The crux: the apparatus that makes a jack-of-all-trades lose to a specialist — so guilds form instead of one generalist sweeping the board. §1 left this here on purpose.

One finite budget, split among machinery

• Each function (uptake, fixation, engulf, secrete-enzyme, swim, buoyancy, adhere, transfer, emit, sensing) needs machinery — the GRN sets an investment level for each.
• All machinery draws on one finite budget: the cell's surface, and the structure to build it. Strictly zero-sum — more of one is less of the rest.
• Built machinery costs maintenance continuously — used or idle (it's structure, §1). A broad toolkit is a standing tax.
• A function's capacity ∝ its allocation, then gated by substrate (Liebig/Droop, §1). Roughly linear in machinery — deliberately not diminishing.

Why the generalist loses — and why it didn't, before

A shared budget alone doesn't beat generalists. Two more ingredients do, and missing either lets the generalist back in:

• Idle machinery still costs. A generalist keeps the whole kit built and pays maintenance on all of it — including the parts it isn't using. (In real cells a broad niche means a large unused proteome and a measurable fitness burden.)
• The world is patchy. In a stable, locally single-resource patch, the specialist that put its whole budget there out-grows the spread generalist. Our world — local fields, diffusion-limited patches, local competition — is exactly that.

The returns-shape guard: the anti-generalist force lives in the maintenance overhead + patchiness, never in diminishing within-function returns — because concave returns reward spreading and would quietly rebuild the generalist. So capacity stays ~linear in machinery.

Two costs, not double-counting: a function carries a standing cost to build and maintain its machinery (this section — the anti-generalist overhead) and a per-use throughput cost when it acts (3.2). That's faithful — enzymes cost to keep and to run.

Grounding: this is textbook proteome + membrane economy — a finite budget split among sectors, idle expression taxed, optimal growth a yield-vs-burden tradeoff, and exactly the constraint that sorts microbes into specialist vs generalist niches. It adds no new currency: machinery is built structure on finite surface, costing maintenance — all from §1. Whether and which guilds emerge stays empirical and tunable (§6: budget size, maintenance rate, patch scale).

What death leaves behind, and how it returns to the living — resolving O1. The 3.2 feeding routes already fix most of it; this pins the physics.

Two forms, with a flux between them

• Death → a particulate particle (POM) where the cell died, carrying its organic C/N (§1: dead carbon still holds its energy).
• POM sinks slowly (tunable — marine snow / the biological pump), and is engulfable (phagotrophic detritivory) by cells in contact.
• POM → DOM two ways: a slow passive hydrolysis (so matter is never locked away, and DOM is always seeded) and a fast enzymatic route (secrete-enzyme, 3.2 — the accelerator, and a public good open to cheaters).
• DOM is dissolved organic C/N: it diffuses — joining §2.4's diffusing species, which until now were only inorganic — and is absorbed by uptake/osmotrophy.
• Mineralization is biotic: a cell that absorbs organic C respires it to inorganic C, and excretes organic N as inorganic N — closing the loop back to the autotrophs. Decomposers earn their keep by returning nutrients.

Conservation across the new pools: total C/N now spans inorganic + DOM + POM + reserves + structure, and the canary must check all of them. Every inefficiency (engulf, uptake, transfer at η<1) routes its un-assimilated matter to detritus or inorganic — never deleted; only energy dissipates. POM resting on the floor still hydrolyses, so the biological pump never permanently buries matter.

Why this softens the old decomposer bottleneck: a decomposer no longer has to chase sinking particles — it secretes enzymes to dissolve nearby POM into DOM that diffuses to it, and passive breakdown seeds DOM regardless. Whether a decomposer guild actually forms stays emergent. Constants → §6: sink rate, passive hydrolysis rate, enzyme rate, DOM diffusion rate.

Checked against the predation-evolution experiments and the digital-evolution platforms, then the conservation seams hardened.

Checked against experiments

• Predation→multicellularity is real and repeated — Boraas (Chlorella), Herron (Chlamydomonas + Paramecium, multicellularity in 2/5 lines within ~750 generations, with measured protection), Becks (rotifer). And the mechanism is ours: predators eat only a narrow size range, so a unit larger than a threshold is immune — exactly engulf-gated-by-colony-effective-size.
• Induced, facultative defense — algae form colonies in response to predators or even their exoproducts, more at higher predator density, reverting when predation lifts. Matches our sense → GRN → adhere path (a prey reading non-kin predator markers and bonding), and it's something to let emerge, not code.
• Coloniality has costs (slower, lower per-cell uptake) — our bond upkeep + reduced per-cell surface — and the outcome is stochastic (it evolved in a fraction of lines), so success is per-run probabilistic, not guaranteed.

Checked against implementations

• Avida / Aevol show that specialization, adaptive radiation, division of labour, host-parasite dynamics and cooperation are all reachable in digital evolution — but their "metabolism" is abstract (logic-task energy; curve-fitting), with no conserved matter, fields, or physical cells. A transferable lesson: decoupling sensors from actuators improved evolvability — which is our 3.1→GRN→3.2 split.
• Our niche is distinctive: conserved matter + reaction-diffusion + DEB metabolism + GRN + physical multicellularity in one model — closer to the PhysiCell / trait-based-ecology lineage. More grounded and mechanistic, at the cost of being heavier with more emergent uncertainty to tune.

Fixes folded in

• H1 — η<1 conserves matter: un-assimilated matter routes to detritus/inorganic; only energy dissipates. §3.4
• H2 — the canary spans the new pools: total C/N includes DOM + POM. §3.4
• H3 — markers & enzymes are massless/catalytic, energy-costed — not tracked matter pools (else they'd leak). §3.1 / §3.2
• H4 — capex vs opex: standing machinery cost (3.3) and per-use cost (3.2) are both intended, not double-counting. §3.3
• H9 — no permanent burial: floor POM still hydrolyses, so the pump isn't a matter sink. §3.4

Logged for §4 / §5 / §6