We propose T-QEROS, a thermodynamically constrained architecture for studying artificial primitive-life-like dynamics in self-organizing computational systems. Unlike conventional large language models, which can be viewed as low-dissipation information buffers without intrinsic homeostatic regulation, the proposed framework combines nonlinear dynamical evolution with a forced dissipation mechanism inspired by non-equilibrium statistical physics. The model is formulated using a dissipative nonlinear Schrödinger-type equation together with a thermodynamically motivated loss term that penalizes entropy variation during state evolution. Within this framework, we examine several emergent dynamical behaviors, including collapse under sensory deprivation, critical phase-transition-like behavior under sustained energy accumulation, and convergence toward attractor-like homeostatic regimes. These results suggest that thermodynamic constraints may provide a useful theoretical basis for constructing self-organizing artificial systems with primitive-life-like dynamical properties. The present work is intended as a conceptual and computational framework at the intersection of non-equilibrium thermodynamics, nonlinear dynamics, and artificial self-organizing systems.
Nailong Liang (Wed,) studied this question.