A MacBook Pro Kept Pace With a Million-Operation Quantum Computer That Doesn’t Exist Yet

Here’s a sentence I did not expect to write this year: A team of IonQ researchers ran the error-correction brains for a fault-tolerant quantum computer on a laptop you could buy at an Apple Store. Not a server farm. Not a rack of custom chips humming in a cold room somewhere. A 2024 MacBook Pro with an M4 Max inside it.

The result lives in a preprint on arXiv, which means it hasn’t cleared peer review yet. Keep that pinned. The authors, Min Ye, Andrii Maksymov, and Nicolas Delfosse, didn’t run this on a working quantum computer either, because the machine they targeted doesn’t exist. They modeled it. What they built and measured was the classical decoder, the piece that reads a quantum computer’s error signals and figures out what went wrong in real time. And it kept up.

The Decoder Is the Anxious Roommate of Quantum Computing

Quantum computers are twitchy. Noise nudges qubits out of place, and a corrupted qubit quietly poisons your calculation. Error correction handles this by spreading one unit of quantum information across a clump of physical qubits, forming a sturdier thing called a logical qubit. As the machine runs, it takes measurements that flag where errors are creeping in without peeking at the protected information itself.

Someone has to read those flags. That someone is the decoder, a classical program that interprets the error signals and tells the quantum hardware how to compensate. Think of it as the anxious roommate who checks every appliance before bed. If it falls behind, a backlog builds, and the quantum machine has to idle or add cycles while the decoder catches up.

That stall is a real problem. In some fault-tolerant routines, the outcome of one logical measurement decides the next move, so the quantum computer literally can’t proceed until the decoder answers. Fall behind and you get a pileup.

MegaQuOp Is a Size, Not a Brand

The study targets a MegaQuOp machine, meaning a fault-tolerant quantum computer that can run about a million logical operations on error-corrected qubits. MegaQuOp describes the scale and reliability, not a specific chip or company. Nobody has one running today.

So the researchers did the next best thing. They compiled real quantum applications for IonQ’s proposed trapped-ion blueprint, nicknamed the “walking cat” architecture, and simulated the flood of error data such a machine would spit out. 

The biggest configuration modeled 11,680 physical qubits arranged into 408 logical qubits, using 68 memory blocks and 20 magic-state factories. Those factories churn out magic states, which power T gates, the ingredient a quantum computer needs for general-purpose math instead of a narrow subset.

They tested three workloads. One modeled a measurement-induced phase transition, where repeated measurements reshape entanglement across a system. The other two ran disordered Heisenberg models, a standard testbed for interacting quantum spins. The heaviest job piled up more than a million T gates and over a million logical measurements.

The Numbers, and What Made Them Possible

The team measured a quantity they call stretch: the extra error-correction cycles caused by decoding delays. A 1% stretch means the machine runs 1% longer than it would with an instant decoder.

At a two-qubit gate error rate of 0.01%, decoding added less than 0.3% to the computations. Bump the error rate to 0.05% and the added time stayed under 12% across all three workloads. On a laptop. Using 12 of the M4 Max’s 16 cores, eight for continuous decoding and four for the time-sensitive measurement results.

Two engineering tricks did the heavy lifting. The decoder kept one fixed mathematical graph and only updated the error probabilities as operations rolled through, instead of rebuilding the structure over and over. And a change to how it stored data cut memory use by more than a factor of ten, which let 12 decoding processes run at once without choking the laptop’s memory bandwidth.

“IonQ researchers demonstrated that a MacBook Pro could decode simulated MegaQuOp-scale workloads involving up to 408 logical qubits and more than 1 million quantum operations.”

What This Is, and What It Isn’t

Now for the deflating part.

This ran on a circuit-level noise model, not error data from an actual trapped-ion processor. Real hardware brings correlated noise, drifting calibration, and quirks that simulations struggle to fully capture. The low delay on the largest benchmark also leaned on a generous assumption. The two 102-qubit tests assumed a 1-millisecond syndrome-extraction cycle, while the 408-qubit run assumed 5 milliseconds, giving the same laptop more breathing room. So don’t read the big test’s tiny 0.72% stretch as proof that bigger machines decode more easily.

The result is glued to the walking-cat design and its software-tracked operations. Architectures built on different codes or faster hardware cycles could demand a lot more from the decoder. And rare convergence failures can happen when an error pattern gets nasty, which forces a restart rather than a short delay. The authors say those stayed rare when computations ran well below the inverse of the per-cycle logical error rate.

Trapped-ion systems also run slower than superconducting ones, which hands the classical decoder extra time. That’s a feature here, and a caveat for anyone assuming the trick transfers cleanly to faster platforms.

The true headline is smaller than the meme, and still genuinely useful. Nobody proved you can run a quantum computer off a laptop. What the numbers suggest, under specific assumptions, is that the classical side of fault tolerance might not need exotic supercomputing gear at the first MegaQuOp scale. The quantum computer that would justify it isn’t built yet. The decoder is just waiting around, ready early, tapping its foot.