Invention as a Structural Operation: Why the Abductive Jump Is a Re-Coordinatization of the Space of Description, Not a Sensory Leap

A prevailing position holds that generative models have mastered induction and are conquering deduction but are structurally incapable of the abductive "jump" from sensory experience to new axioms, which is said to require embodied simulation and physically grounded world models. This paper argues that the jump is misdescribed. Invention is a typed structural operation on a representation – a change of the description itself – realized by four atomic moves: adjoining a dimension, rescaling or requantizing, re-topologizing, and gluing across domains. These moves are selected by a functional combining intrinsic structural signals (description length of the axioms, symmetry, and closure) with an often-weak empirical term that can act during generation rather than only downstream. Within representational systems whose compatibility predicate is decidable, the operator set is enumerable and the selection is computable, so the generative act is mechanizable, as a running artifact that performs and classifies cross-domain structural recombination demonstrates. The account is bounded: outside decidable domains the operator types describe the moves without an enumerability or completeness claim. Evolution – invention without a mind – runs on the same four moves, showing structural invention to be substrate- and cognition-independent. Includes zharnikov-2026bg-invention-as-structure.yaml (Paper Spec v0.1.0) – a machine-readable specification of the paper's claims, assumptions, and dependencies. The paper's full machine-first bundle (the SPINE claim/dependency graph, the ONTOLOGY term module, and the reproducible companion demonstration under `code/`) lives in the public repository at https://github.com/spectralbranding/meaningfulness-papers/tree/main/invention-as-structure; see https://github.com/spectralbranding/paper-spec for the standard. This PDF is generated programmatically from that machine-first source under a research-as-repository model.