This paper bridges the prior law stack into engineering by mapping the four PIEC constraints onto AI alignment architectures. It analyzes multiple architecture classes in terms of feasibility, channel irreducibility, sustained exogenous access, and corrective authenticity, and introduces the ideas of corrective ecology, amplification over substitution, and the complexity-starvation threshold as bridge-level engineering concepts. The focus is not on proving a single favored architecture, but on showing how physical correction constraints translate into architectural requirements and vulnerabilities.
Taylor Prather (Mon,) studied this question.