Lossy by Construction: A Substrate Audit of Computational Physics is a methodological audit of the numerical substrate on which modern computational physics is performed. The paper does not argue against continuum physics as theory. It asks a narrower and more practical question: when continuum equations are evaluated on finite-precision binary hardware, what is actually being computed, and how fully is that numerical situation being described? The paper develops four linked observations: First, many methods presented as discrete lattice physics are better understood as carrier-field methods, where the lattice provides an addressing structure while the physical content is carried by continuous per-site quantities. Second, mainstream equations contain transcendental constants such as π and e, which no finite substrate can represent exactly and must be borrowed approximately. Third, the floating-point substrate itself is finite, lossy and operand-dependent, yet its precision budget is rarely treated as a first-class part of the reported result. Fourth, one standard infeasibility objection to more substrate-honest computation is overstated: the addressability requirements for substrate-scale bookkeeping are large but not absurd. The paper does not propose alternative physics. Its conclusion is methodological. Precision bookkeeping, representational limits, borrowed constants and substrate-dependent error should be treated as reportable parts of a numerical physics result rather than as silent background conditions absorbed by the machine.
ALAN BALL (Tue,) studied this question.