A qubit -- the quantum equivalent of a bit -- is the most fragile object in the known universe of computing. Not fragile in the way a hard drive is fragile. Fragile in the way that a soap bubble is fragile, except that the bubble is made of physics rather than film, and anything -- a stray photon, a vibration in the room, a fluctuation in the magnetic field -- will pop it. For decades, the central challenge of quantum computing has been this: how do you build a computer out of objects that are destroyed by the act of measuring them? In April 2026, two independent research groups published answers that make the problem, for the first time, tractable.
In classical computing, a bit is either 0 or 1. It stays that way until you change it. Errors happen, but they are rare, detectable, and correctable. In quantum computing, a qubit can be in a superposition of 0 and 1 simultaneously -- this is the property that makes quantum computers potentially far more powerful than classical ones for certain classes of problems. But superposition is not a stable state. It is an inherently fragile quantum condition that collapses into a definite 0 or 1 the moment the qubit interacts with its environment.
This process is called decoherence, and it happens on timescales of microseconds to milliseconds -- fast enough that the qubit forgets what it was computing before the computation is complete. The problem compounds: quantum error correction requires multiple physical qubits to encode a single logical qubit, and measuring those physical qubits to check for errors also risks disturbing the state you are trying to protect. The core dilemma is that the tools needed to diagnose errors are themselves sources of errors. Until recently, the time required to characterize how a quantum processor was failing was measured in hours. By the time the diagnosis was complete, the failure modes had changed.
A quantum computer with a 0.1% error rate per gate, running a circuit with 1,000 gates, has less than a 37% chance of producing a correct answer. For the algorithms that make quantum computing useful -- Shor's algorithm for factoring, Grover's algorithm for search, quantum chemistry simulations -- circuits require millions of gates. Without error correction that reduces effective error rates to below 1 in a billion, fault-tolerant quantum computing is impossible.
Shadow tomography is the solution that makes it possible to characterize a quantum state without destroying it. The key insight is statistical: instead of directly measuring the qubit (which collapses the superposition), you run the actual computation in parallel with a "shadow" version -- a lightweight sampled procedure that extracts statistical information about the state without fully observing it. Each shadow measurement captures partial information.
Across many measurements, those partial observations build up a statistical picture of the full quantum state -- and of the errors accumulating within it. The technique was first proposed theoretically, but a paper published in Nature in April 2026 demonstrated it working in practice, detecting errors in milliseconds -- approximately 100 times faster than previous characterization methods. Separately, researchers at the University of Innsbruck, working with partners in Sydney and Waterloo, published a technique called cycle error reconstruction that makes individual qubit errors visible during logical operations. Rather than diagnosing errors after a computation fails, cycle error reconstruction monitors the error landscape in real time as the computation runs.
A third breakthrough, published simultaneously, introduced PAEMS -- a Predictive Adaptive Error Model for Superconducting qubits -- that demonstrated error correlation reductions of 19.5-fold, 9.3-fold, and 5.2-fold in timelike, spacelike, and spacetime dimensions respectively on IBM quantum processors. Together, these three results represent a step-change in the ability to understand and control quantum hardware in real time.
The goal of quantum error correction is fault-tolerant quantum computing: machines that can correct errors faster than they accumulate, running circuits of arbitrary length and complexity. The threshold theorem in quantum computing states that if the error rate per gate falls below a certain threshold (typically around 1%), fault-tolerant computation becomes possible using surface codes -- a form of quantum error correction in which logical qubits are encoded in a 2D lattice of physical qubits, with errors detected by measuring neighboring qubits without disturbing the logical state.
IBM, Google, IonQ, and Quantinuum have all been racing toward this threshold. Google claimed to cross it with Willow in late 2024. IBM's latest processors (Heron and beyond) have demonstrated two-qubit gate fidelities above 99.5%. The challenge has shifted from individual gate fidelity to system-level error management -- tracking and correcting errors across thousands of physical qubits running millions of gates simultaneously. Shadow tomography and cycle error reconstruction address this system-level challenge. They give engineers something they have never had before: real-time, high-resolution visibility into how errors are distributed across the processor. The IBM PAEMS results are particularly significant because they demonstrate this working on real production hardware, not laboratory prototypes.
The applications of fault-tolerant quantum computing that are most discussed -- breaking RSA encryption, simulating molecular chemistry from first principles, solving certain optimization problems exponentially faster than classical computers -- require circuits of a scale that current noisy hardware cannot execute. To break a 2048-bit RSA key using Shor's algorithm requires roughly 4,000 logical qubits running millions of error-corrected gates. Current state-of-the-art machines have hundreds of physical qubits and error rates that would cause such a computation to fail immediately.
The error correction breakthroughs of April 2026 do not deliver fault-tolerant quantum computing. They make the engineering path to it clearer and faster. Shadow tomography reduces the characterization time for hardware failures from hours to milliseconds, which accelerates the design iteration cycle. Cycle error reconstruction identifies which specific physical errors most affect logical gate performance, allowing hardware teams to target improvements precisely. PAEMS provides a predictive model that anticipates how errors will change over time, enabling proactive correction rather than reactive diagnosis. The combination is a significant tightening of the feedback loop between quantum hardware designers and the machines they build.
The systems that will eventually run Shor's algorithm are not yet built. But the tools being developed now -- the ability to watch a quantum computer fail in real time, at millisecond resolution, and understand exactly why -- are the tools that will build them.
© 2026 Lisa Pedrosa · lisapedrosa.com
All articles cited to primary institutional or peer-reviewed sources
Buy me a coffee