Classical computing asks every wire to be a clean yes or no. Quantum computing lets a wire be a weighted blend of both until you look — and that single change reshapes what computation can be. Here is the working vocabulary.
A classical bit is the smallest decision a machine can hold: it is either 0 or 1, and at any instant it is exactly one of them. Every photo, model, and spreadsheet is built from billions of these committed choices.
A qubit refuses to commit early. Its state is written |ψ⟩ = α|0⟩ + β|1⟩, where α and β are complex numbers carrying both a size and a phase. The qubit genuinely is that combination — not secretly one value we haven't checked, but a real superposed state with structure of its own.
One definite value at all times. Reading it changes nothing.
A weighted blend of 0 and 1 with phase. Reading it forces a choice.
The weights must satisfy |α|² + |β|² = 1. Those squared magnitudes are the probabilities of measuring 0 or 1, and probabilities have to total 1 — so a qubit's state always lives on the surface of a sphere, never inside it.
Put n qubits together and the system can occupy all 2ⁿ classical patterns simultaneously, each with its own complex amplitude. Three hundred qubits would carry more amplitudes than there are atoms in the observable universe.
The catch — and it is the whole game — is that you cannot read out all of those amplitudes. Measurement collapses everything to a single n-bit string. So a quantum algorithm is never "try every answer and print them." It is the art of arranging interference so the wrong answers cancel and the right ones reinforce, before you look.
Amplitudes can be negative or complex, so two paths to the same outcome can add (louder) or subtract (silence). Classical probabilities only ever add. That subtraction is the resource quantum computers actually spend.
A single qubit's state, ignoring an irrelevant overall phase, is fully described by two angles. That makes it a point on the surface of a sphere — the Bloch sphere. The north pole is |0⟩, the south pole is |1⟩, and the equator holds the even superpositions, separated only by their phase.
Every single-qubit operation is just a rotation of this sphere. That is the quiet payoff of the picture: abstract complex algebra becomes "turn the globe," and the next instrument lets you do exactly that by hand.
Gates are the verbs of quantum computing. Each is a reversible operation — mathematically a unitary matrix — that rotates the state without ever destroying information. Here are the workhorses you will meet everywhere.
A small kit — say {H, T, CNOT} — can approximate any quantum operation to arbitrary precision. You don't need infinitely many gate types, just the right handful, exactly as classical logic reduces to NAND.
Apply a Hadamard then a CNOT to two fresh qubits and you reach the Bell state (|00⟩ + |11⟩)/√2. Now neither qubit has a state of its own — only the pair does. Measure one and find 0, and the other is instantly 0 too; find 1, and its partner is 1, no matter how far apart they sit.
This is not a hidden message being sent — you can't use it to signal faster than light, because each qubit alone looks like random noise. What entanglement buys computation is correlation structure: a way to wire qubits together so operations on one ripple through the whole register at once.
Every gate so far is reversible. Measurement is the one door that only swings one way. The instant you read a qubit, its superposition collapses to a single definite value — 0 or 1 — with probability equal to that outcome's squared amplitude. The rest of the state is gone.
So the rhythm of a quantum program is always the same: prepare a superposition, interfere with gates so the answer dominates, measure once. Often you repeat the whole run many times and read the answer from the distribution of outcomes, the way you'd learn a coin's bias by flipping it repeatedly.
Quantum computing carries a lot of hype. Worth separating the textbook physics, which is not in dispute, from the engineering and the open claims, which very much are. And the line moves: one item below just crossed from frontier to settled.
Historically, adding qubits added error sources — more parts to go wrong — so bigger codes performed worse. "Below threshold" means the opposite finally happened in hardware: enlarge the code and the logical qubit gets better, exponentially. That is the threshold theorem confirmed on a real chip — the precondition for everything in the frontier column eventually scaling. The proof of work: