Abstract
Artificial-life research has largely treated silicon as an inert execution layer and located life-like dynamics inside software. This paper argues that this assumption misreads the substrate. We make three observations about conventional CMOS systems. First, MOSFET 1/f noise arises from distributed charge-trap dynamics and exhibits a spectral form associated with self-organised criticality, motivating investigation of silicon as a critical-like substrate. Second, the cache hierarchy creates a physical latency gradient that can support selection-like differential persistence among self-reinforcing memory-topology structures without a software-defined fitness function. Third, DRAM operates as a dissipative non-equilibrium memory substrate, sustained by continuous refresh, nonlinear readout, and intrinsic fluctuation. Taken together, these observations suggest that conventional silicon already possesses structural preconditions relevant to emergent life-like organisation. We propose substrate-coupled artificial life as a research programme: not the simulation of life on silicon, but the experimental investigation of life-like organisation as a property of silicon substrates under controlled non-equilibrium operation. The work is theoretical and pre-experimental, and the framework is offered for falsification.