CODA-GTLA: General Theory of Layered Authority - Six jobs, Not one machine: Obstruction, Standing, Transport, Comparison, Shadow, Leftover

"Six Desks, Not One Machine: Why a Leftover Class Is Not Standing." In modern computing and agent architectures, it is tempting to conflate evaluation with authority: to compute a loss residual, transport a representation, match an embedding, or inspect a coarse projection, and treat that mathematical leftover as if it granted a license to act. It does not. Glue is not a grant, a blurry photo is not the original, and an algebraic leftover is neither a decision nor an origin story. CODA-GTLA defines the coordinating spine across the Calculus of Obstruction, Descent, and Authority (CODA). Rather than absorbing its companion papers into an amorphous heuristic, GTLA establishes strict structural boundaries between six independent operations: Obstruction (Φ₁, Γ₂)): Measuring linear repair deficits in finite cochain complexes. Standing: Demanding an inhabited pipeline terminal backed by seven exact halt guards. Transport: Tracking collisions across commuting observation squares. Comparison: Pullback gluing of overlapping relational spans. Shadow: Enforcing fiber constancy under non-injective coarse projections. Leftover / Anomalon: Retaining cokernel residuals (x − Cx) as typed remnants rather than discarded noise. Supported by Lean 4 constructor separation and exact rational Python test suites, GTLA demonstrates how a finite evaluator traverses these layers in strict sequence, ensuring operational authority is granted only by verified evidence rather than computational leftovers.