BlueQubit's CPU/GPU simulators are exact statevectors — they store the full 2^n amplitude vector, so 34–36 qubits is the hard ceiling regardless of circuit structure. We take a different approach: MPS/tensor-network representation that scales with entanglement rather than qubit count. For highly entangled circuits we're strictly worse. For structured circuits (most variational, most shallow) we can go much larger. Complementary tools really.