Fate of Reality

August 31, 2026

Quantum Just Crossed a Line

IBM and the University of Chicago ran a 15-minute computation that classical computers can't practically reproduce, and published the receipts.

The facts:

  • IBM and University of Chicago researchers completed a quantum computation in about 15 minutes that leading classical methods can't practically reproduce (ScienceDaily / IBM)
  • The system ran 70 error-corrected "logical" qubits, one of the largest demonstrations of logical quantum computing reported so far (ScienceDaily)
  • Inside the run: 2,415 logical two-qubit operations and 468 logical T gates (ScienceDaily)
  • Effective logical error rates came in 10 times lower than the underlying physical error rates (ScienceDaily)
  • The old benchmark, random circuit sampling, had a built-in flaw: once the task beats classical computers, classical computers also can't check the answer (ScienceDaily)
  • This team designed a more structured circuit specifically so the result could be verified. The paper is called "Sampling hard circuits with verifiably high fidelity" (ScienceDaily)
  • Circuits and results are public on the Quantum Advantage Tracker (ScienceDaily)
  • IBM's framing: they're now "firmly in the quantum advantage era" (ScienceDaily)

Quantum news is easy to skip. It's always some lab getting a handful of qubits to behave for a fraction of a second longer, the press release calls it a revolution, nothing changes. So "quantum advantage era" invites the obvious response: sure you are.

Then there's the number 70, and it's worth staying for.

Seventy logical qubits is a different animal from seventy qubits. A logical qubit is quantum information wrapped in error correction so the noise that normally wrecks these machines in microseconds doesn't get through. Error correction has been the wall this whole time. Making qubits is a solved problem. Making them reliable isn't. And here's a system where the corrected error rates came in ten times better than the raw hardware underneath. The wrapping held while the machine did thousands of real operations.

The part that actually makes this one credible is duller than any of that. It's the verification.

The old way to claim quantum supremacy was a benchmark where a quantum computer generates patterns too complicated for a classical computer to reproduce. Convenient problem with that: too complicated to reproduce means too complicated to check. You had to take the lab's word for it. This team built their circuits differently so the answer could be verified, put it in the paper title, and posted the circuits publicly where rivals can attack them.

Now, nobody's password got cracked this week. This was a benchmark built to prove capability, not a drug discovered or a material designed, and it's not clear the verification trick scales to bigger computations. "Quantum advantage era" is still an IBM sentence about specialized tasks, not about problems anyone will pay to solve. That gap is probably still years wide.

But think about how machine learning looked ten years ago. Benchmarks and papers, benchmarks and papers, and then all at once it was in everything. If quantum follows anything like that curve, the milestone people point back to won't be a flashy demo. It'll be a boring one, where the errors got corrected and the results got published somewhere checkable.

So did quantum cross the line this week? IBM says yes. The honest answer is that they crossed a line, verifiably, in public, and whether it's the line depends on questions nobody can answer yet.

Source: ScienceDaily, on the IBM / University of Chicago demonstration.