The Quiet Calculus: Inference, Adaptive Memory, and Conditional Physics in ZoraASI and Theory of Everything Research (version 4.0)

A source-grounded scientific synthesis that extracts 92 page-referenced principles from the Dune campaign novelization The Quiet Calculus, Book I: First Cohort, connects them to versioned ZoraASI and Theory of Everything research, specifies twelve falsifiable prediction cards, and reports bounded local computational tests run under rules fixed in advance. Version 4.0 tests targeted repairs on newly held-out material. A gated adaptive-memory repair passes its pre-set criteria on 1,100 new simulated paths. Two evidence-judgment repairs fail their pre-set tests (21/36 vs 26/36; 18/24 vs 19/24). The second removed the targeted error, calling undecidable claims false, but over-abstained and misread negatively worded claims. The version 3 results are preserved unchanged. Contents: the paper (PDF), full LaTeX source, the version 3 and version 4 evidence bundles (frozen protocols, raw model calls, audits), and versions 1.0 to 3.0 of the paper. Status: preprint, not peer reviewed. Prepared with AI assistance. The evidence is computational (software checks, local language-model behavior on AI-written fictional cases, and simulations), with no laboratory or human-subject data and no physical validation of a Theory of Everything.