Qubits lose their quantum superposition states as they interact with their environment---effectively getting "measured" by their surroundings. This is called d

Qubits lose their quantum superposition states as they interact with their environment---effectively getting "measured" by their surroundings. This is called decoherence, and is the central engineering obstacle to building useful QCs. The longer a qubit lasts while maintaining its quantum coherence, and (especially) the more operations you can do on it while keeping it coherent, the better that qubit's quality. In practice, we will never be able to build qubits of perfect quality (i.e., ones that maintain their coherence forever except when we deliberately measure them). In the 1990s, some people thought this would be a fatal obstacle to scaling up QCs. But then two closely-related discoveries, called quantum error-correction and quantum fault-tolerance, dramatically changed the picture. These discoveries showed that, even if your physical qubits fall short of perfection, as long as they have a high enough quality, you can glom a bunch of them together into a single "logical qubit"---that is, a qubit that lives in the collective state of multiple physical qubits, and that can still be recovered even if any small number of those physical qubits lose their coherence. Furthermore, one can do an arbitrarily long quantum computation on these encoded (logical) qubits, continuously monitoring to see which of the physical qubits have suffered errors and correcting those errors (but not monitoring in a way that would collapse the logical qubits!). In this way, one can in principle build a reliable quantum computer out of unreliable parts---a generalization of John von Neumann's famous discovery from the 1950s, which showed that the same was true of classical computation.