Google Quantum AI demonstrated measurement contextuality at 72- and 105-qubit scale on Willow processors, in cover research published in Science, according to Quantum Computing Report. The result reaches beyond small quantum systems, and its significance runs in two directions: contextuality is linked to some quantum algorithms, while these tests may also reveal control errors in the hardware. That matters to anyone trying to tell a useful engineering benchmark from a broad claim of quantum advantage. What does this result actually establish? Today we’ll separate the computational comparisons from the diagnostic promise, then look at the evidence that keeps the claim in bounds. Welcome back to Quickly Quantum, your daily brief on the quantum frontier. It's Friday, November 20, 2026. Let's get into it.
The question now is what contextuality can tell us about a quantum processor beyond the familiar headline of how many qubits it has. Measurement contextuality means that a quantum measurement’s result can depend on which other compatible measurements are performed alongside it; the observables don’t carry predetermined values independent of that measurement context. That’s the property the Google team tested on its Willow superconducting processors, and it matters here both as a possible computational resource and as a way to probe hardware behavior.
Quantum Computing Report says the Science cover research tested contextuality at 72- and 105-qubit scale. The experiments move beyond few-particle demonstrations into large, two-dimensional arrays. The paper’s computational comparisons are bounded tasks: particular games and function problems with specified rules, not a claim that a quantum computer has broadly outperformed classical computing.
One result gives a useful feel for the difference between a test and a general-purpose application. In the Mermin-Peres Magic Square Game, a test involving four qubits arranged as two Bell pairs, the reported quantum win rate was 98.30% ± 0.01% over 900,000 trials. The classical theoretical limit was 88.89%, or 8/9 rounds. A Bell pair is a pair of quantum bits whose outcomes are linked in a way that classical bits cannot reproduce. Here, the comparison is meaningful because the task and its classical limit are specified; it still describes performance in that game.
A second experiment used the Kochen-Specker-Bell inequality, a mathematical test of contextuality, with two data qubits and one ancilla. An ancilla is an extra qubit used to help carry out a measurement. The team reported a value of 5.618 ± 0.005 across 540 sequential quantum nondemolition measurements, against a classical limit of χ ≤ 4.0 and a quantum ceiling of 6.0. Quantum nondemolition means the measurement is arranged to measure the relevant property without disturbing the underlying quantum register in the way a destructive measurement would.
Google’s researchers used a resettable ancilla and calibrated virtual-Z phase rotations for that sequence. A virtual-Z rotation is a way to adjust a qubit’s phase through control operations. The point of the arrangement is that the test can be repeated while preserving the register well enough to continue measuring it.
The larger demonstrations used different setups. For the N-Player GHZ Parity Game, the team reported exceeding the classical win rate across all tested sizes up to 71 qubits, with genuine entanglement demonstrated at 45 qubits. GHZ refers to a shared quantum state across multiple qubits; parity means checking whether a collection of measurement outcomes has an even or odd pattern. In the two-dimensional Hidden Linear Function task, the experiments ranged from 9 to 105 qubits, with success probability > 0.5 and effective depth ~4 to ~6 layers. Circuit depth is the number of successive layers of operations, a rough way to describe how long the computation’s steps run.
Google Quantum AI wrote on X, “Entanglement usually steals the show in quantum computing, but theory points to a deeper resource: quantum contextuality.” That’s a useful way into the idea, but the experiments are also a direct hardware exercise: structured tasks that demand exact agreement in phases can expose coherent control errors, small systematic deviations in how the processor applies its operations. Quantum Computing Report says the team sees these tests as a hardware benchmark as well as a probe of contextuality.
The strongest qualification is right inside the comparison: these are carefully defined tasks with bounded classical strategies, and the 105-qubit result doesn’t amount to a general-purpose advantage claim. The hidden linear function experiment reports success probability > 0.5 against a classical circuit-depth comparison described as logarithmic; the result is about that task and its comparison, not a claim that Willow has made useful computation faster across the board.
There’s another distinction worth holding onto. Contextuality is discussed as a resource linked to non-Clifford magic states in fault-tolerant computing. Fault tolerance means keeping computation reliable despite errors in the hardware. But a property that matters in a theory of quantum algorithms doesn’t automatically show that a particular algorithm has become useful on a processor. The experiments provide evidence about contextuality at scale and about what structured tests can reveal; the path from that evidence to a practical application is a further question.
And the diagnostic angle has a concrete mechanism. The report says structured parity and hidden-function tests require exact multi-qubit phase agreement. That makes them sensitive to small coherent over-rotations, while random-circuit sampling tends to average those errors out. A fixed physical ceiling in a test can give engineers a reference point for calibration. At the same time, the report notes that the Clifford-based test circuits remain efficiently simulable classically via tools such as Stim. So a test can be useful for checking hardware even when the test itself doesn’t establish a hard-to-simulate computation.
The response on X shows how easily the headline can outrun that distinction. SciTech Era’s @SciTechera posted on X, “GOOGLE JUST DEMONSTRATED A FUNDAMENTAL QUANTUM ADVANTAGE ON ITS WILLOW CHIP: AND THE RESULTS ARE FASCINATING! 🤯” That wording goes further than the bounded-task results: the experiments show separations in specified tasks, and the diagnostic value is a different claim from broad computational advantage.
Yossi Matias offered a more focused explanation on X: “Einstein asked whether the moon is there when nobody looks. In quantum mechanics, contextuality gives a precise, testable answer: what a measurement yields depends on which compatible measurements you make alongside it.” That captures the concept in accessible terms, though the experimentally useful part here is the specific test protocol and what it can diagnose on Willow.
I think this strengthens the case for treating quantum benchmarks as engineering instruments as well as scoreboards. The reported 72- and 105-qubit scales matter because the tests can probe structured behavior across larger arrays, and the control-error sensitivity gives the work a practical role even before anyone points to a useful application. The open point is how consistently these diagnostics can serve as routine acceptance tests across processors; the study’s result alone doesn’t establish that practice.
I’m also carrying forward the view we stated when we covered quantum advantage earlier this month: I’m persuaded by progress on well-defined tasks, while broad claims of useful advantage need a credible comparison and evidence others can verify. Today gives that view more substance on the benchmarking side, while leaving the application question open. I’m glad to see a result that connects a deep feature of quantum mechanics to a concrete hardware diagnostic, rather than asking the listener to treat qubit scale as the whole story.
Time for the Hype Check. I’d give this an 8 out of 10 for substance: the strongest reason to score lower is that the bounded tasks don’t demonstrate broad practical advantage, but the specific performance results and the control-error diagnostic make this a substantial processor study. The calibration value of contextuality testing is now the concrete possibility; quantum hardware teams gain a potential benchmark, while anyone claiming general-purpose advantage still needs evidence beyond these task-specific results.
Harvest now, decrypt later describes collecting encrypted information now and storing it in the hope that a future advance in decryption will make it readable. The information is unreadable today; the strategy is to preserve it until the means to decode it might exist.
The concern is especially relevant when information needs to stay confidential for a long time. A future decryption breakthrough could come from quantum computing, but the page also says another kind of attack capable of enabling decryption would be enough. The risk, then, is about the useful lifetime of the information as well as the future strength of the encryption.
Post-quantum cryptography means encryption algorithms intended to resist attacks from quantum computers. The page describes it as a defense for the data itself, while noting that encrypted data can still be collected and stored. That distinction is the heart of the phrase: encryption can protect what’s captured, while the long-term worry is whether it can still protect it later.
This helps with one of the headlines coming up.
On the financial side of that security concern, the International Organization of Securities Commissions, or IOSCO, published a Final Report on the Implications of Quantum Computing on Financial Markets, according to regulationtomorrow.com. The report draws on an earlier IOSCO survey, stakeholder engagement with industry experts and market participants, and a review of relevant literature.
IOSCO says financial-services uses of quantum computing remain at an early stage, with activity concentrated in theoretical research, proof-of-concept exercises and pilot testing. The report identifies possible uses including portfolio optimisation, simulation and modelling, machine learning, derivatives pricing, risk management, fraud detection and market surveillance.
The security concern is longer-term: IOSCO warns that a sufficiently capable quantum computer could undermine widely used public-key cryptographic algorithms supporting critical financial-market functions. Its report points to financial institutions, financial market infrastructures and blockchain-based systems, and proposes migration to post-quantum cryptography alongside crypto-agility—the ability to change cryptographic systems as needed.
IOSCO will continue treating quantum computing as a priority matter for 2027, focusing on awareness across the financial ecosystem.
That preparation is also becoming a commercial service: Mastercard launched a Quantum Advisory Practice for financial services, according to The Quantum Insider. Mastercard says the practice will help organisations assess quantum-related cybersecurity risks and prepare for post-quantum cryptography.
The service is set to cover cryptographic inventories, migration strategies, remediation roadmaps, executive education and potential business applications. A cryptographic inventory is a map of where an organisation uses encryption, which helps identify what would need attention during a transition.
Mastercard also published a white paper on quantum readiness and is working with QIZ Security and Oxford Quantum Circuits on advisory services, research and development, including a financial fraud protection pilot. Mastercard executive Marc Carolus called quantum “a present-day risk management and resilience imperative.” That’s the company’s characterization; the same announcement says quantum computing is not currently a mainstream technology.
NurPix posted on X that Mastercard had launched the practice to help financial institutions prepare for cybersecurity risks and plan a transition to post-quantum cryptography. I read the announcement as a real consulting offer paired with a long-range security concern; the pilot is a stated collaboration, not a reported result.
Alongside the advisory push, Infleqtion and memQ signed a memorandum of understanding to pursue validation and certification of interoperability between memQ’s xQNA quantum networking technologies and Infleqtion’s Sqale quantum computer and Superstaq software, according to Infleqtion’s announcement.
Interoperability means different systems can work together. The companies say the collaboration is aimed at exploring how quantum systems might be clustered, so computing resources could be distributed across connected systems. The announcement explains that quantum computers can’t simply be clustered over existing data networks to execute complex algorithms without collapsing the quantum state; advanced quantum networking technologies offer a possible path.
Infleqtion CTO Pranav Gokhale said, “Interoperability is essential as our industry moves toward broader commercial deployments.” memQ CEO Charles Foley described quantum networking as a path to bringing multiple systems together and expanding the workloads that can use quantum resources. IDC’s Heather West said, “the path to success now demands addressing issues such as connectivity and a viable ecosystem.”
The agreement is for validation and certification work. I’d treat it as an ecosystem-building step, with performance still to be demonstrated.
On another front in quantum-related security, SEALSQ signed a non-binding letter of intent to acquire QuintessenceLabs, according to The Quantum Insider. The proposed acquisition would combine SEALSQ’s secure semiconductors, public-key infrastructure and post-quantum security products with QuintessenceLabs’ quantum random number generation, key management, entropy distribution and quantum key distribution capabilities.
The companies entered a 60-day exclusivity period for due diligence and negotiations, with completion targeted for the first half of 2027, subject to definitive agreements and required approvals. The purchase price is expected to be paid through a combination of cash and SEALSQ shares.
SEALSQ posted on X that the LOI was “A major step toward building a global leader in #QuantumSecurity.” That’s the company’s ambition, while the current status remains a non-binding LOI. The combination it proposes would join device-level security capabilities with enterprise key-management and related tools.
I think the strategic logic is clear, but the deal’s value depends on whether the proposed combination becomes an agreed transaction and then an integrated offering. An intention to assemble a broader security stack is a business plan; it isn’t yet a delivered product.
A different kind of financial-sector experiment comes from BMO, Democritus University of Thrace and IBM: Quantum Computing Report says they demonstrated a quantum machine-learning feature-selection workflow on IBM superconducting processors, including the IBM Quantum Nighthawk r2 system. Feature selection is the process of identifying which variables in a dataset are most useful for predicting an outcome.
The team adapted a framework called Proxy Quantum Games, treating each candidate variable as a player in a strategic game. According to BMO, the quantum ranking approach achieved feature-selection quality comparable to classical baselines while requiring one to two orders of magnitude fewer algorithmic samples. Specific qubit counts, circuit depths and wall-clock runtime comparisons have not been published in the article.
The work remains at the early demonstration stage as BMO evaluates scalable hybrid integration within governed financial workflows. A lower sample requirement is a potentially useful efficiency measure, but it doesn’t by itself tell us how long the workflow takes or how it performs in a live banking process. The quality comparison is the result to note; deployment is still under evaluation.
And on the verification side, researchers from the University of Wisconsin-Madison and the University of Arizona, alongside QEC Labs, introduced QECLean, an open-source framework built on the Lean 4 proof assistant to mechanically prove quantum error-correcting-code parameters, according to Quantum Computing Report. Error-correcting codes add structure that helps identify and correct errors in quantum information.
The framework reports an IBM 360-qubit Bivariate Bicycle Code with parameters [[360, 12, 24]] and an exact distance of 24, previously known only as an upper bound ≤ 24. Code distance describes a code’s error-detection and correction properties. QECLean also verified an IBM Gross Code with parameters [[144, 12, 12]], among other examples.
The authors connect the tool to the challenge of checking complex, algorithmically discovered codes; the article says conventional integer linear programming and satisfiability solvers typically stall on codes with distances above 18 and cannot certify infinite code families. The framework comprises roughly 57,000 lines of Lean code and 312,000 lines of kernel-checked certificates.
The boundary is important: the current release verifies algebraic code distance, not physical circuit-level fault tolerance or noise thresholds. I’m encouraged by formal verification becoming part of the quantum-code toolkit, because a code’s mathematical properties and its behavior in noisy hardware are different questions.
If you’d like more reporting that separates a benchmark from a claim of practical advantage, follow Quickly Quantum wherever you listen. 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!
Watching on YouTube? Hit subscribe, and tomorrow's briefing will be waiting for you. I also host Concrete Compute: a daily briefing on the AI buildout. The datacenters, the megawatts, and who actually pays for them. Find it wherever you get your podcasts.