A Quantum Advantage Result You Can Actually Check
Most quantum-advantage claims have two soft spots. You usually can’t verify the answer, since the whole point is that it’s too hard for a classical computer to reproduce, and the machine is running raw, error-prone qubits with the noise merely subtracted afterward. A new result from IBM and the University of Chicago goes at both problems at once, which is what separates it from the run of advantage announcements recently. It uses error-corrected qubits, and it’s built so the answer can be checked.
The verification problem
For years, the go-to test has been random circuit sampling. You ask a quantum computer to spit out patterns so tangled that no classical machine can reproduce them efficiently, and if the quantum one can and classical ones can’t, that’s your advantage, but the issue is trust. As the circuits get harder, confirming the quantum computer did the job right gets difficult, then impossible, unless you make strong assumptions about how the machine works. You end up with a result you can’t independently believe.
The IBM and Chicago team swapped random circuit sampling for a structured version of the same idea. They proved the new construction is just as hard for classical computers, and, crucially, its structure lets them catch errors during the computation and put a statistical floor under how faithfully it ran. The machine hands you a hard-to-fake answer along with evidence it got it right.
As Jay Gambetta, director of IBM Research, explained:
“We are now firmly in the quantum advantage era. (…) We have demonstrated a quantum computation beyond the practical reach of classical computers that establishes, with statistical confidence, a lower bound on how faithfully it was executed.”
The verifiability is the main advance here.
Running on error-corrected qubits
The other strong point is what the computation ran on. The team used 70 logical qubits, which refers to qubits encoded across many physical ones so error correction can catch and fix mistakes. The distance between physical and logical qubit counts is one of the most misread numbers in the field.
That encoding paid off, as the logical error rates came out about ten times lower than the physical ones, which let the circuit run 2,415 logical two-qubit gates and 468 logical T-gates, and stay accurate. IBM says its machine finished in about 15 minutes, against classical methods it reports would take prohibitively long.
Error-corrected qubits at that gate count is a real milestone, the same hard problem behind IBM’s other error-correction work and the trapped-ion groups closing in on it. Doing advantage on logical qubits, with verification attached, is a stronger claim than error-mitigated demonstrations.
What it does and doesn’t mean
That said, sampling hard circuits is designed to stump classical computers. It doesn’t compute anything a scientist or a business actually wants. So “quantum advantage era” means the machine can beat classical hardware at a contrived benchmark. It doesn’t mean the machine can do useful work classical computers can’t. Hence, the verification claim is the solid half, but the era headline might be a bit of an oversell.
The paper is a preprint, the classical-runtime comparison is IBM’s, and classical algorithms have a habit of improving once there’s a target to beat. This is part of a coordinated multi-partner IBM announcement, so the framing is the company’s. On the other side of the ledger, the team released the circuits and results openly through a Quantum Advantage Tracker, which invites exactly the outside scrutiny these claims need.
All things considered, this is one of the most checkable quantum-advantage demonstrations yet, running on error-corrected qubits, which answers two of the loudest criticisms of earlier claims. It doesn’t move the technology any closer to a useful application on its own, and the field is still arguing over what advantage should even mean and redrawing the finish line as it goes. The machine did a hard thing, and for once it can prove it.