Quantum Technology

Intuition A classical bit is a switch with two positions. A qubit is a direction on a sphere: the north pole is 0, the south pole is 1, and every other point is a blend of the two. That sounds like room for enormous amounts of information — but the moment you read it, it snaps back to one pole or the other, and what comes out is still a single classical bit.

Mechanism A qubit is written Describing n qubits takes 2ⁿ complex numbers; at three hundred qubits that is more amplitudes than there are atoms in the visible universe. But this pile is not there for you to read, it is there for you to steer: the entire craft of a quantum algorithm is arranging for amplitudes leading to wrong answers to cancel in pairs while the right answer adds up.

The counterintuitive part "A quantum computer tries every possibility at once" is the most stubborn misconception in the field. If that were true it would have crushed every search problem long ago. The branches do evolve together — but you get to carry away exactly one result, and you don't get to choose it. The answer has to be hidden inside an interference pattern, and that is brutally hard to design: forty years on, the algorithms with a proven exponential speedup can be counted on your fingers.

Intuition Superposition is fragile not because it "turns into some other state" but because phase leaks away. One stray photon, one flicker of magnetic field, and the environment is entangled with your qubit — it has effectively recorded whether that qubit leans 0 or 1, and the interference pattern collapses into ordinary randomness. This is the mechanism from the issue on measurement and interpretation showing up as an engineering problem: nobody has to be watching; the environment is the observer.

Mechanism Engineers track two times: T₁, how long the energy takes to drain away, and T₂, how long the phase stays in step — the second is shorter and the one that hurts. Superconducting qubits have T₂ in the tens-to-hundreds of microseconds, gates take tens of nanoseconds, so the natural budget is a few thousand operations: nowhere near enough for a useful algorithm. The way out is error correction — but you cannot store three copies and take a vote, because copying an unknown state is forbidden outright. Instead a single logical qubit is smeared across a patch of entangled physical qubits, and you only measure the syndrome: whether neighbouring qubits agree. The syndrome says nothing about the state itself, yet it pinpoints where an error struck. The threshold theorem then supplies a guarantee: as long as the physical error rate sits below a threshold (about 1% for the surface code), the bigger you make the patch, the faster the logical error rate falls.

The counterintuitive part Crossing the threshold is not a manifesto; it was measured at the end of 2024. On one superconducting chip, 101 physical qubits were encoded into a single distance-7 logical qubit with a logical error rate of 0.143% per correction cycle, living 2.4 times longer than the best single physical qubit on the same chip — the first time more correction made things better instead of worse. But read the other half of the ledger: that bought one logical qubit, and factoring-scale tasks want thousands.

Intuition The first quantum technologies out of the lab were not computers but sensors and links: they never need hundreds of qubits held in entanglement, only a few quantum states kept quiet enough. And the fragility is the selling point — something exquisitely sensitive to its surroundings makes a superb probe.

Mechanism Communication rests on one prohibition: the no-cloning theorem. Quantum evolution is linear, "copy an arbitrary unknown state" is not, so that photocopier cannot be built. Eavesdropping therefore has to disturb the signal, and comparing the error rate on a sacrificed slice of key reveals whether anyone was listening.
Sensing rests on how violently phase responds to an external field. With N independent particles the phase precision is The examples that have landed are concrete: optical lattice clocks stable at the 10⁻¹⁸ level (less than a second of drift over the age of the universe); gravitational-wave detectors injecting squeezed light to push quantum noise down and see further; and in 2017 the Micius satellite distributed entangled photon pairs to two ground stations 1200 km apart.

The counterintuitive part "Quantum encryption is absolutely secure" needs discounting. What is proven is that the protocol is secure with ideal hardware, and real attacks never go after protocols — they go after devices: early on, people blinded detectors with bright light so they would report whatever the eavesdropper wanted. Add that it only solves key distribution and needs dedicated fibre or satellites, and you can see why national cryptography agencies push post-quantum cryptography instead — maths that classical machines can run and quantum machines still cannot break, deployable as a software update.

Intuition Three questions will screen out most quantum promises: is there a proven speedup? how does the data get in? how many logical qubits does it need, and how many orders of magnitude away are we? The list of applications that clear all three is short.

Mechanism Only two families have solid evidence of exponential speedup. One is simulating quantum systems themselves — molecular electronic structure, strongly correlated materials; Feynman's 1982 reasoning was plain enough: use quantum things to compute quantum things and the cost stops exploding. The other is problems with hidden periodic structure, factoring being the emblem. Quadratic speedups come from search: finding one item among N unsorted takes N classical steps and √N quantum ones. That margin is easily eaten by error-correction overhead. The most over-sold category is "quantum machine learning": many claimed exponential speedups turn out to be matched by classical algorithms under the same data-access assumptions. And loading N classical numbers into a quantum state costs about N steps by itself, which vaporizes an exponential advantage on the spot.

The counterintuitive part A "quantum supremacy" demo is not the same as being useful. The famous 2019 demonstration ran random circuit sampling — a task chosen precisely because it is hard for classical computers and has no known use; classical algorithms have been narrowing the gap ever since. The criterion worth watching is: beating the best classical method, reproducibly, on something people already wanted done. That day has not arrived.
But don't overcorrect either. The physical qubits estimated to break 2048-bit RSA fell from twenty million in 2019 to under a million in 2025; error correction has just crossed threshold; sensing and timekeeping are in service. This is not a bubble — it is a field moving on a decade clock, packaged by a narrative running on a quarterly one.