From a Connectome to a Digital Fly Executable Neural Models, Sensorimotor Closure, and Source-Shadow Cartography in TCGS-SEQUENTION

Public demonstrations of fruit-fly connectomes controlling virtual bodies and video games have made the passage from anatomical reconstruction to executable behaviour unusually visible. This article examines that passage within the source-first architecture of \TCGS: complete non-temporal Counterspace, physically real three-dimensional Shadow carriers, gauge-equivalent ordering labels, and a source-grounded Extrinsic Constitutive Law distinct from its mathematical representations. The central analytical object is not a wiring matrix alone but a closed experimental model comprising anatomical constraints, neuronal dynamics, sensory encoding, motor decoding, body--environment transitions, initial conditions, and a validation protocol. Primary connectomic and modelling studies establish genuine predictive achievements; developer documentation identifies additional engineering choices in public demonstrations. These evidentiary classes are evaluated separately. We derive explicit counterexamples showing that an unchanged weighted graph can support different stability regimes, that changing only a motor interface can alter closed-loop stability, and that identical external responses can conceal different internal dynamics. Eliminating an unmodelled motor subsystem produces a memory-dependent effective law rather than an automatically justified neuron-to-action shortcut. A local sensitivity analysis further identifies when decoder freedom can absorb apparent evidence for connectivity parameters. These results motivate closure-conditioned attribution: a claim about the contribution of biological wiring is assessed only after its dynamical, interface, environmental, and measurement conditions have been declared. The proposed validation programme separates exact representational equivalence, numerical convergence, physiological uncertainty, and genuine physical interventions; combines frozen and refitted topology controls; and requires held-out neural and behavioural predictions. The typed factorization $\pi_\sigma^{\rm bio}=\mathsf R_\sigma^{\rm bio}\circ\pi_\sigma^{\rm phys}$ keeps biological observables within the physical-readout chain rather than introducing a separate source channel. A claim-level verdict combines a validity gate with joint uncertainty sets to distinguish not evaluable, rejected, criterion met, and inconclusive outcomes for compatibility, contribution, specificity, invariance, and biological transfer. Small reproducible analytical calculations illustrate the distinctions without being presented as a full-connectome benchmark. The resulting account preserves the scientific value of connectomic simulation while rejecting the identification of task performance, biological embodiment, source identity, and experiential access as one achievement.