Latest / Quickly Quantum / August 17, 2026 — Frolov vs. Microsoft: Quantum's Replication Crisis
Transcript
- How much of the last decade's biggest quantum computing breakthroughs would actually survive someone trying to replicate them? That's the question underneath our main story today — a New Scientist feature revisiting a replication study taking direct aim at the physics behind Microsoft's Majorana bet. Before we get there, in the headlines: a Brazilian team says they can build any single-qubit gate with just one shaped pulse of light, Rice University physicists found they can flip a magnetic signal with nothing but a pinch of mechanical strain, and a German lab wants to rewrite the rulebook on how we even measure quantum advantage in the first place. Welcome back to Quickly Quantum, your daily brief on the quantum frontier. It's Monday, August 17, 2026. Let's get into it. Fraunhofer researchers in Germany just published two papers arguing that how we measure quantum advantage — the point where a quantum computer definitively beats a classical one — is too idealized to mean much. This comes via the institute's own release, so keep in mind it's their read on their own work, but the underlying critique stands on its own. The first paper looks at quantum chemistry, where molecules are usually modeled as perfectly closed systems, sealed off from everything around them. Real molecules aren't like that — they leak energy, they relax, they're constantly interacting with their environment. Florentin Reiter, who heads Fraunhofer's Quantum Systems group, put it this way: 'The exciting question is not just whether quantum computers can outperform classical computers, but when, why, and under what conditions.' The second paper tackles the QAOA algorithm, often pitched for portfolio optimization, and proposes testing for favorable scaling as problems grow bigger, instead of declaring victory on one convenient problem size. Now, this is a proposal, not a new result — it only matters if labs and vendors actually adopt the tougher standard. But it's a fitting warm-up for our main story: another corner of the field asking whether quantum computing's claims of success are actually earning it. A team at the Universidade Federal de São Carlos in Brazil just simplified something that's usually a headache in quantum hardware: building a single-qubit gate, the basic operation that flips or rotates a qubit's state. Normally that takes a sequence of carefully timed pulses. This team showed that a single shaped pulse of light — modulated in both frequency and amplitude — can, in theory, produce any single-qubit gate you want. The trick is mathematical: instead of numerical optimization, which is basically trial and error run by a computer, they inverted the equation describing how a qubit evolves and got a closed-form formula for the pulse directly — an analytical recipe instead of a black box. The one simplifying assumption is the rotating-wave approximation, a standard tool for describing how light interacts with a two-level system. This is single-source, simulation-stage work — nobody's run it on real hardware yet, so noise robustness and real-device fidelity remain open questions. But if it holds up experimentally, it could simplify the control hardware across several qubit types. Over at Rice University, Pengcheng Dai's team found something that could matter for spintronics — carrying information with electron spin instead of charge. They studied an altermagnet, a material called manganese telluride that breaks time-reversal symmetry — its physics runs differently forward versus backward — while having almost no net magnetization. That's great in theory and a mess in practice, because the material splits into magnetic domains pointing every which way, muddying the signal. Their fix: physically stretch it. Strain merged the competing domains and let them clearly measure the anomalous Hall effect, a sideways voltage that reveals the internal magnetic structure. The number that stood out: tuning strain by just one percent flipped the sign of that signal entirely, and the team calculates that's comparable to changing temperature by around 150 degrees — a far more practical knob than heat or a giant magnet. This is materials physics, not a qubit milestone, and near-term device applications are still speculative. But a strain-tunable magnetic switch working at everyday temperatures is exactly the building block spintronics has been chasing. One more, quick and theoretical: researchers at the University of Connecticut, along with a collaborator at the Tata Institute of Fundamental Research in India, built a mathematical theory that predicts something called prethermal plateaus — long stretches where a strongly interacting quantum system seems to pause on its way to equilibrium before settling down. That pause matters for hardware, because it's essentially a window for how long a quantum simulator or processor can hold onto useful information before noise wins. The team's model, which they call a fragmented eigenstate thermalization hypothesis, predicts both how high that plateau sits and how long it lasts — a calculation that was previously out of reach. It applies to systems with long-range interactions, the kind you find in trapped-ion and Rydberg-atom experiments. This is purely theoretical for now; it needs experimental confirmation on real hardware before it changes how anyone designs a chip. But if the predictions hold, it's a genuinely useful map for how long today's noisy quantum processors have before thermal noise takes over. Our main story today: quantum computing's replication problem, and whether the industry's incentives are built to catch it or bury it. New Scientist has a feature out revisiting a saga that's been simmering since January, and it goes straight at the physics underneath Microsoft's flagship quantum chip. Here's the setup. A string of headline experiments claimed to have found something called Majorana zero modes — exotic particle-like states inside hybrid superconductor-semiconductor devices that, in theory, could be stitched together into a topological qubit, a qubit design where noise-resistance is supposedly built into the physics itself rather than bolted on afterward with extra error correction. That's the bet Microsoft has staked its Majorana 1 chip and its whole 'Topological Core' roadmap on. The problem: a team led by Sergey Frolov at the University of Pittsburgh, working with collaborators in Minnesota and Grenoble, spent years trying to reproduce those milestone results. And when they wrote it up, they found the dramatic 'smoking gun' signals — the kind of clean, unmistakable-looking patterns that made headlines in the first place — dissolve into much more mundane superconducting effects once you look closely, effects that trace back to fine-tuning quirks in complicated samples rather than the exotic physics being claimed. Now here's the part that should bother you almost as much as the science itself: Frolov's team submitted this replication paper back in September of 2023. It got rejected, repeatedly, by the same journals that had published the original breakthrough claims — with editors reportedly citing a lack of novelty, or suggesting the field had simply moved on, as if checking whether a milestone is real stops mattering once everyone's stopped talking about it. It finally ran in Science on January 8th of this year, and the team says the only reason it got through at all is that they bundled several separate replication attempts into one paper specifically to force the issue past that bias. This isn't Frolov's first rodeo, either. He's the same physicist who initiated a high-profile 2021 Nature retraction over an earlier Majorana claim, and he's been openly skeptical of Microsoft's Majorana 1 rollout, calling it a project resting on physics that hasn't actually been established. So there's a track record here — this is a recurring voice in one of quantum physics's most contested corners, not someone taking a first shot at a hot topic. Now, to be clear about what this paper is and isn't saying: Frolov's team frames this as a call for better data-sharing norms rather than an accusation of fraud — their own language is that dramatic smoking-gun patterns can signal real, important effects in topological condensed matter physics, but in these specific cases they trace back to mundane fine-tuning in complex samples rather than anything exotic. Microsoft, for its part, hasn't conceded that the underlying physics is invalid — the company has kept publishing follow-up Majorana results since. The debate over whether topological qubits are achievable remains genuinely unsettled, not resolved against the approach. And this isn't a single outlet running with a hot take, either — Phys.org and ScienceDaily both covered the University of Pittsburgh's own writeup, and SciTechDaily picked it up too, so we've got more than one source independently confirming the details here. So here's the deeper read, and it's less about whether Majorana particles are real and more about how the field decides what to believe in the first place. Think about the incentive structure Frolov's team ran into: they spent years doing careful replication work, and reviewers at multiple journals turned it down for not being 'novel' enough — as if confirming or debunking a foundational claim is somehow less valuable than making the original splashy claim in the first place. In quantum computing specifically, that bias runs straight into an industry raising real money and building hardware roadmaps off claims that, by this account, hadn't been adequately checked. It's the same discomfort with rigor we heard in the Fraunhofer benchmark papers earlier — a different corner of quantum computing, but the same instinct to declare victory before the checking is done. For a company the size of Microsoft, a wobble in the underlying physics means a hardware roadmap resting on unconfirmed ground-floor claims until someone finishes the check. Here's why this matters if you're not a physicist: quantum funding, roadmap slides, and 'we hit a milestone' headlines all trace back to claims like these — precise, hard-to-verify physics results announced well before the field has settled whether they're real. When verification takes three years and a workaround just to get printed, the public conversation runs on the unverified version for that entire stretch. That pipeline — fast hype, slow verification — extends well past Microsoft's chip; qubit counts, coherence times, and quantum-advantage claims across the industry move through the same pattern, which is exactly the gap the Fraunhofer papers are trying to close from the other direction. And that's the question I keep coming back to: how much of the last decade of quantum hardware milestone news — the smoking-gun result, the record qubit count, the 'first demonstration of X' — got the same kind of scrutiny Frolov's team eventually forced through, and how much just stuck, because nobody had the incentive to go check? The process story here is the one I find most damning, honestly. A replication paper submitted in 2023 shouldn't need three years and a workaround — bundling multiple replication attempts together just to get past reviewers who'd rather move on to the next headline — to see daylight. If that's happening in one of the most closely watched corners of quantum computing, with a heavily funded roadmap riding on the outcome, it raises an obvious question about every other breakthrough story this field has produced that hasn't gotten the Frolov treatment yet. What would actually move this forward is another replication attempt, done in the open, with data shared well enough that outside groups can pull it apart themselves before it ever reaches a press release. That's essentially the title of Frolov's own paper: data sharing helps avoid smoking-gun claims of topological milestones. Until that becomes normal practice instead of a fight to get published, this doesn't really resolve — it just moves to the next hardware announcement. Time for the Hype Check. I'm putting this one at a 7. The paper itself is solid — peer-reviewed in Science, corroborated by multiple outlets, written by the same physicist who forced a 2021 Nature retraction, doing the same homework again with a bigger sample and a workaround to even get published. That's substance, and this is a case where the process itself is part of the story — a journal system that took years to greenlight a check on a marquee industry claim. What keeps it from an 8 or 9 is that the 'replication crisis' framing risks running ahead of what's actually been shown — this doesn't prove Majorana zero modes don't exist, and it doesn't prove Microsoft's chip is built on nothing. It proves a specific set of prior smoking-gun claims don't hold up to closer scrutiny, and that the review process made it needlessly hard to say so. Important, well-sourced, but don't let anyone on either side of this tell you it's more settled than that. So here's the open question I don't have an answer to yet: if the field can't get a straightforward replication paper published quickly even when the stakes are this high, what would it actually take for the rest of us to trust the next smoking-gun claim when it comes? If today's episode left you a little more skeptical of the next 'breakthrough' headline, good — that's kind of the job here. If you want that same skepticism delivered fresh every weekday, follow Quickly Quantum wherever you're listening, and if this one landed, send it to the person in your life who still thinks quantum computers are just faster regular computers. This has been Quickly Quantum, an AI-voiced podcast, created and built by a real human using today's cutting-edge technology. Nothing you heard on this show is financial advice. I'm Brian Lampert, and I'll catch you all tomorrow — take care!