SNT-LIFE: A Surviving Digital Organism Governed by LLM-Driven Operator Selection in a Predator-Prey Ecology

We introduce SNT-LIFE v2, a digital ecology in which a digital organism survives under survival-constrained computation within a predator-prey ecology. The organism is governed by an operator algebra—seven structured cognitive operators, each with a distinct energy homeostasis cost—selected via LLM executive control that dynamically reweights operators in response to environmental threat. Survival is formalized as the Survival Viability Condition (SVC), a constraint-satisfaction objective that replaces reward maximization: the organism must maintain metabolic energy above a death threshold rather than optimize a scalar reward. In a 2D simulation with food scarcity and delayed rewards, SVC-driven organisms survive 63. 4 13. 4 steps versus 47. 1 17. 9 steps for rule-based baselines (+34. 6\%, p < 10^-11). A stress-test analysis reveals real-time adaptive behavior: the Phase Reverser (evasion) operator weight increases from 0. 08 to 0. 50 within one step of crisis onset. Ablation studies confirm the functional contribution of each operator. The framework yields five falsifiable predictions with explicit null hypotheses and clarifies the distinction between reward as environmental energy influx versus reward as optimization objective. SNT-LIFE v2 demonstrates that survival pressure—not reward design—is the source of adaptive behavior in a digital ecology.