Qubit count is the headline number and the wrong one. A device of fifty to a few hundred qubits passes the point where brute-force classical simulation is feasible, but with two-qubit gate error rates above a tenth of a percent and no error correction, the executable circuit is capped at roughly the reciprocal of the error rate in gates, so noise rather than scale sets the computational ceiling. Reading the near-term landscape through that constraint reorders it: hybrid variational schemes that keep circuits shallow are the plausible near-term paradigm, exponential speedups that depend on loading classical data or preparing a particular input state may lose their advantage to the loading cost, and a claimed speedup can evaporate entirely when the quantum algorithm teaches someone how to build a classical one. The constraint also cuts the other way, since a circuit made more noise resilient may become easier to simulate classically.
John Preskill (Institute for Quantum Information and Matter and Walter Burke Institute for Theoretical Physics, Caltech), arXiv 1801.00862v3, quant-ph with cond-mat.str-el cross-list, dated 30 July 2…