What Quantum Can and Cannot Do Today, an Honest List

• Run small programs on real hardware. You can write a circuit, send it to a real quantum processor over the cloud, and get results back. This works right now, and you can do it for free. The programs are small and the results are noisy, but the machines are real, not simulated.
• Solve certain optimization problems on annealers. D-Wave's annealers can take real optimization problems today. Whether they beat a good classical solver on problems that matter is still genuinely debated, but the door is open and usable now.
• Serve as research and learning instruments. For scientists studying quantum systems, and for anyone learning the field by running real experiments, today's machines are already valuable. A lot of honest current value is here.
• Occasionally beat classical computers on a contrived task. A few times, a quantum machine has performed a narrow, carefully chosen task faster than the best classical supercomputer. These demonstrations are real milestones, but the tasks were chosen to favor the quantum machine and are not useful applications. They prove the hardware is doing something genuinely quantum, not that your business problem is ready.

• Break the encryption that protects the internet. The famous algorithm that threatens today's public-key cryptography needs far more high-quality, error-corrected qubits than anyone has. It is a real future risk worth preparing for, not a present capability.
• Design drugs or new materials at production scale. Simulating molecules is one of the most promising long-term uses, and small demonstrations exist, but useful, better-than-classical chemistry at scale is not here yet.
• Speed up your everyday software. Quantum computers are not faster classical computers. They help only with specific problem shapes, and for the vast majority of computing, a classical machine is and will remain the right tool.
• Deliver reliable, fault-tolerant computation. Today's machines are in the noisy era. Run a circuit too long and the errors pile up until the answer is meaningless. The fault-tolerant machines that the big promises depend on do not exist yet.

Between "can now" and "cannot yet" is a real and active zone: researchers hunting for a genuinely useful problem where a noisy near-term machine beats the best classical method. Finding a clear, practical win there would be a big deal. As of now, that decisive, useful, better-than-classical result on a real-world problem has not landed, and anyone who tells you it has is either selling something or has moved the goalposts.

Hold the two columns in your head and re-sort every claim you read into them. The test is simple: is this something you could do on a real machine this afternoon, or is it something that depends on hardware nobody has built yet? Both are legitimate to talk about. The dishonest move, and the common one, is to blur them. Quantum is genuinely useful today for a narrow set of things and genuinely promising for a much larger set later. Keeping the two straight is not cynicism. It is the only way to invest your attention in the field without getting burned by its hype cycle.