Each of the 150 organisms carries a synthetic genome g = (armor, metabolism, neural, locomotion), every gene in [0,1]. Bio-augmentation isn't free: after every mutation the genome is rescaled so the four genes sum to a fixed budget B = 2.2 — pumping one trait necessarily starves the others, the same energetic trade-off real organisms face.
fitness(g) = 0.5 + 1.3·armor·hazard + 1.3·metabolism·scarcity
+ 0.4·neural + 0.4·locomotion
death chance = clamp(hazard·(1-armor) + scarcity·(1-metabolism) - 0.35, 0, 0.95)
Each generation: (1) every organism rolls against its death chance — high radiation kills the unarmored, scarcity starves the metabolically inefficient; (2) survivors are ranked by fitness and the top selection pressure fraction becomes the breeding pool; (3) 150 offspring are produced by picking two random parents from that pool, averaging their genomes (crossover) and adding Gaussian noise of size σ (mutation rate) to each gene before re-normalising to the budget.
- Radiation hazard / Resource scarcity — the environment. They set both the death chance and which genes fitness rewards, so the population should visibly drift toward armor under radiation and toward metabolism under scarcity.
- Mutation rate σ — variance injected each generation; too low and the population stalls, too high and beneficial genomes get destroyed as fast as they appear.
- Selection pressure — how narrow the breeding pool is; a tiny pool converges fast but crashes genetic diversity, a wide one drifts slowly.
- Auto-Evolve — steps one generation roughly every 900 ms so you can watch the mean genome (the bars) converge live.
This is the same directed-evolution / evolution-strategy loop (fitness → selection → crossover → mutation) used to engineer stress-tolerant microbes and optimise protein variants in real synthetic-biology pipelines — here scaled up to a whole augmented organism.