quantum interference

Think about ripples on a pond. When two waves meet, they add where their crests line up and cancel where a crest meets a trough. Quantum interference is the same idea, but the things doing the adding and cancelling are amplitudes: the numbers a quantum computer carries for each possible outcome. The catch that makes them quantum is that amplitudes can be negative or complex, not just positive like a probability. So two paths leading to the same answer can reinforce each other, or they can wipe each other out.

This cancelling and reinforcing is the actual engine of quantum algorithms. You do not get speed by having the machine quietly try every answer at once and hand you the right one. Instead, you arrange the circuit so that, by the time you measure, the amplitudes pointing at wrong answers have largely cancelled and the amplitudes pointing at the right answer have piled up. The probability of reading any outcome is the square of its amplitude (the Born rule), so steering amplitude toward the answer is what makes the answer likely to appear. Build the interference well and you win; build it badly and you just get noise.

This is also why quantum computers are not magic general-purpose accelerators. Interference only helps when the problem has structure you can exploit to make the wrong-answer paths cancel. Shor's algorithm uses the quantum Fourier transform to make periods interfere into view, which is why it threatens RSA; Grover's search uses interference too, but only buys a quadratic (square-root) speedup, not an exponential one. For many everyday problems there is no clever interference pattern to set up, so a quantum computer offers no advantage at all.