Mosaic gigantism in otodontid sharks: trophic convergence onto lamnid apex predators without full locomotor convergence

The extinct macropredatory shark Otodus megalodon is reconstructed at 16–24 m and 30 to over 100 t, but most estimates rest on isolated, toothless vertebral centra and unverified cross-species scaling. We integrated tooth-based, vertebra-based, and volumetric size proxies, keeping one-dimensional length separate from three-dimensional mass, and coupled them to a dual-envelope vertebral mechanics model and extant-anchored locomotor allometries. Tooth-inclusive length proxies converge on adult lengths of about 18–20 m; we adopt an 18.5 m, approximately 43 t working model under a slender, fineness-ratio-6.06 form. The often-cited 24.3 m, approximately 94 t animal rests on an untested cross-specimen extrapolation and is treated as hypothetical. Mass depends on body form, not length: at 18.5 m the stockiest published curve approaches 97 t, whereas at 24.3 m a density–fineness box caps the body at 122–135 t, excluding a stocky approximately 200 t form. Axial safety factors permit cruising at every size, but maximal fast-start performance declines with size: approximately 43 t is the practical ceiling for an ambush-and-short-sprint, enlarged-white-shark ecology, while a 94 t animal stays below a safety factor of unity even at the maximal mineralization-compensation envelope. Denticle morphology, vertebral architecture, and zinc isotopes indicate slow cruising (0.9–1.4 m s^-1), brief ambush bursts (6–7 m s^-1), and scavenging rather than sustained mako-like pursuit. O. megalodon thus exemplifies mosaic gigantism: trophic convergence onto lamnid apex-predator function without complete locomotor convergence. This preprint has not undergone peer review. All fossil measurements are from published, publicly catalogued specimens (institutional catalogue numbers are listed in Table 1). The processed data, parametric body mesh, and analysis code are available from the corresponding author on reasonable request and will be made publicly accessible upon publication. An AI assistant (Doubao, developed by ByteDance) was used solely for English-language editing and writing/debugging Python code, as declared in the manuscript; it did not generate hypotheses, data, analyses, results, or conclusions.