Part 5 · classical competition
A quantum-advantage claim cannot be built on the failure of one classical method. Here, three classical views tell very different stories.
The global MPS warning
For the full 61-qubit depth-2 state, graph-aware MPS runs remain strongly inconsistent between χ = 64 and χ = 128. That says the tested global representation is not converged. It does not say that every useful output is hard.
The Majorana truncation warning
At depth one, exact Majorana propagation is an excellent independent check. At depth two, a weight-32 truncation has a maximum error of about 0.380 on the tested local panel. Weight 64 exceeds the explicit 200,000-term boundary. This route is not competitive for the selected task.
The lightcone resolution
Exact backwards cones solve 41 observables, while a local graph-aware MPS converges the remaining one. The strongest classical method is therefore neither the global MPS nor the truncated Majorana solver, but a task-specific hybrid of exact cones and one local tensor network.
Sources and reproducibility
- Bose et al., A Spin Entanglement Witness for Quantum Gravity
- Marletto & Vedral, Gravitationally Induced Entanglement
- Sabín, Digital quantum simulation of quantum gravitational entanglement
- Pastawski et al., Holographic quantum error-correcting codes
- IBM Quantum documentation: ZNE and PEA
- Public code, numerical reports and full 42-value table on GitHub
Project status: 14 August 2026. Numerical values come from the frozen public research artifacts; claim boundaries are deliberately preserved.