The Theory Machine: A Minimal Operational Epistemology for AI and Science

Many AI and scientific workflows still treat “theory” as a narrative artifact: coherent explanations are promoted into policy, products, or papers without explicit falsifiers, replayable evidence, stable measurement stamps, or verification-budget accounting. We propose the Theory Machine (THM), a minimal operational epistemology that compiles hypotheses into auditable Theory Cards—scope-stamped, falsifier-first, receipt-carrying, and probe-ready—before any promotion occurs. Each Theory Card makes explicit what the hypothesis claims, where it applies, what it predicts, what would falsify it, and which receipts and low-cost probes are required for replayable evaluation. This compilation step converts narrative explanations into operational objects whose promotability can be gated under a fail-closed discipline (HOLD/YELLOW/GREEN) and verified via independent replay, rather than inferred from coherence, authority, or benchmark scores. We position THM as a subsystem inside an Evolution Machine (EM): Solvers generate candidate hypotheses, THM compiles them into operational objects, Mathines execute probes and enforce promotion gates, and Learners extract reusable failure motifs and repairs. Our contribution is intentionally scoped: we demonstrate THM as an executable compiler and replay harness, and we specify an empirical validation roadmap and success criteria for reaching GREEN under independent replay across multiple seeds. We do not claim world-level truth or domain performance prior to such runs.