Post-Deterministic Distributed Systems: A New Foundation for Trustworthy Autonomous Infrastructure introduces Post-Deterministic Distributed Systems (PDDS) as a research and engineering model for coordinating heterogeneous infrastructure in which deterministic services, stochastic models, autonomous agents, policy engines, and human-mediated actors coexist. Classical distributed systems theory has traditionally assumed that correct participants execute protocol-specified behavior with stable, deterministic semantics. This paper argues that autonomous reasoning engines and model-driven agents challenge the universality of that assumption. In such systems, correct participants may produce divergent reasoning paths, distinct operational traces, and heterogeneous internal representations while still reaching semantically equivalent and safe outcomes. The paper frames classical distributed computing as the zero-ambiguity special case of a broader participant-general model. It introduces semantic coherence and admissible behavior as foundations for correctness beyond deterministic transition equivalence. It also outlines five architectural pillars of post-deterministic infrastructure: Protocol-Driven Development, Verifiable Agentic Infrastructure, Autonomous State Control Planes, Semantic Quorum Assurance, and Epistemic State Replication. The work further defines a preliminary taxonomy of post-deterministic failure classes, including semantic drift, correlated reasoning failure, intent loss, evidence fabrication, unsafe delegation, policy-violating autonomy, epistemic divergence, and context amnesia. The goal is to establish a foundation for trustworthy autonomous infrastructure whose behavior remains auditable, bounded by policy, and semantically certifiable. This record provides a timestamped archival version of the manuscript.
He et al. (2026) studied this question.