Qubit
The quantum equivalent of a bit. A bit is off or on; a qubit holds a direction, and measuring it returns just one bit.
Watch for A qubit count alone says almost nothing. Ask whether they mean physical or logical.
Full lesson →
Jargon decoder
Quantum 101
Every term you will hit in an announcement, in plain language, with a note on how it tends to get stretched. If you want the concepts in order instead, start with the lessons.
Basics
4 termsPhysical qubit
An actual circuit on actual hardware. Noisy: every operation has a real chance of going wrong.
Watch for This is the number in most headlines, because it is the biggest one available.
Full lesson →Superposition
A qubit state that combines 0 and 1 with a weight on each. The weights can be negative, which is what makes interference possible.
Watch for "Both at once" and "tries every answer simultaneously" are wrong and lead people to expect the wrong machine.
Full lesson →Entanglement
Two or more qubits sharing a state, so their measurement outcomes correlate perfectly even when separated.
Watch for It does not transmit information. Any claim of faster-than-light signalling is wrong.
Full lesson →Error correction
6 termsLogical qubit
One reliable qubit assembled from many noisy physical ones, with error correction running underneath. The kind an algorithm can actually use.
Watch for If a release gives no logical count, that absence is usually the story.
Full lesson →Error correction
Spreading one logical qubit across many physical ones and measuring only their relationships, so mistakes are found without reading the data.
Watch for "Error mitigation" is a different, weaker thing and the words get swapped.
Full lesson →Surface code
The most studied error-correcting layout, using a two-dimensional grid of qubits. Robust, but expensive in overhead.
Watch for Overhead estimates vary hugely with hardware quality; a single ratio quoted without conditions is a guess.
Full lesson →Threshold
The hardware error rate below which adding more physical qubits makes a logical qubit better rather than worse.
Watch for "Below threshold" is a genuine milestone. Check whether it was demonstrated or projected.
Full lesson →Fault tolerance
Running a computation correctly for as long as needed despite continuous errors. The goal the field is organised around.
Watch for "Fault-tolerant" gets applied to prototypes that demonstrate one ingredient of it.
Magic state
A specially prepared state needed for the gates error correction cannot perform directly. Expensive to produce.
Watch for Magic state distillation dominates the cost of a real machine and is often left out of qubit-count estimates.
Hardware
5 termsDecoherence
The loss of a qubit's state to its environment. The central engineering problem of the field.
Watch for Longer coherence is good, but it is meaningless without the gate error rate alongside it.
Full lesson →Coherence time (T1, T2)
How long a qubit holds its state. T1 is energy loss; T2 is loss of phase relationship.
Watch for Quoted in microseconds and easy to cherry-pick from the best qubit on a chip.
Superconducting qubit
Qubits made from superconducting circuits, run near absolute zero. Used by IBM, Google and Rigetti.
Watch for Fast gates, short coherence. Every architecture trades one against the other.
Full lesson →Trapped ion
Qubits made from individual atoms held in electromagnetic fields. Used by IonQ and Quantinuum.
Watch for Excellent fidelity and connectivity, slower gates. Not directly comparable to superconducting counts.
Neutral atom
Qubits made from atoms held by laser tweezers, rearrangeable into different layouts. Used by QuEra and Pasqal.
Watch for Large arrays are easier here, so raw atom counts can outrun useful qubit counts.
Benchmarks
4 termsGate fidelity
How accurately one operation is performed, as a percentage. 99.9% means one error in a thousand.
Watch for Two-qubit fidelity is the hard number. Single-qubit fidelity is much easier and is sometimes quoted instead.
Two-qubit gate error
The error rate of an operation between two qubits. The figure that most limits how deep a circuit can go.
Watch for The most useful single number in a hardware announcement, and the most often omitted.
Quantum volume
A single figure combining qubit count, connectivity and error rates. Higher is better.
Watch for Vendor-defined, so it is comparable between machines only when measured the same way.
#AQ (algorithmic qubits)
IonQ's metric for how many qubits are usable in a real algorithm after errors.
Watch for Company-specific. Do not compare it directly against another vendor's qubit count.
Eras
1 termNISQ
Noisy Intermediate-Scale Quantum: today's machines, too big to simulate easily and too noisy to error-correct.
Watch for The term implies usefulness is close. Whether NISQ machines can do anything commercially useful is still unsettled.
Claims
2 termsQuantum advantage
A quantum computer doing something faster than any classical method, on some task.
Watch for Several claimed advantages were later matched by better classical algorithms. Ask what the classical baseline was.
Full lesson →Quantum supremacy
The older term for the same idea, usually on a contrived task chosen to be hard classically.
Watch for The task is often useless by design. Supremacy is not the same as usefulness.
Full lesson →Algorithms
2 termsShor's algorithm
A quantum algorithm that factors large numbers efficiently, which would break RSA.
Watch for It needs a fault-tolerant machine far beyond anything built. Demonstrations factor tiny numbers.
Full lesson →Post-quantum cryptography
Classical encryption designed to resist quantum attack. Standardised by NIST and already being deployed.
Watch for This is the actual response to Shor, and it does not require a quantum computer.
Full lesson →