What Does a Quantum Computer Actually Do?

You'll often see quantum computers described as trying all possible answers to a problem simultaneously, then picking the right one. That framing is popular because it's intuitive, and it's wrong in a way that matters. If a quantum computer really did compute every possible answer in parallel, then just read one out at measurement, you'd get a random answer, not necessarily the right one. Superposition alone doesn't hand you the correct result on a plate.

What a quantum algorithm actually does is more precise: it sets up a superposition across many possible states, then applies a carefully designed sequence of quantum gates so that the different paths through that superposition interfere with each other, like overlapping ripples in water. The sequence is engineered so that paths leading to wrong answers cancel each other out (destructive interference) and paths leading to the right answer reinforce each other (constructive interference). Only then is the qubit measured, and by that point the odds are stacked heavily toward reading out the correct answer.

This is why writing a good quantum algorithm is genuinely hard, and why only a handful of problems have known quantum algorithms that beat the best classical approach: designing that interference pattern correctly, for a specific problem, is the actual intellectual work. It's also why "just throw more qubits at it" doesn't automatically make a computation faster; the algorithm has to be built for the hardware to make interference work in your favor.