Can this framework compute a positivity bound rather than cite one?

2.8284271 against 2√2 = 2.8284271, error 2×10⁻¹⁰. The entire model is two constraints: the moment matrix over {1, A₀, A₁, B₀, B₁} is positive semidefinite with unit diagonal. No Hilbert space, no states, no operators.

Classical correlations reach 2, the convex hull of deterministic strategies. Pure no-signalling would permit 4. Quantum sits at 2√2, and where it sits follows from positivity alone.

The point was to validate a table, not to learn the number. The four bound-producing mechanisms went in untested; codifying an unchecked list into a restructure is the error this project keeps finding elsewhere. The same machinery scales to the other listed instances — the sphere-packing LP bound and the conformal bootstrap are the same shape with a larger matrix.