Quantum computing, without the hype and without the sneer

A quantum computer does not do ordinary computing faster. It does a small number of specific computations in a fundamentally different way, and for a handful of problems that difference is exponential. For everything else it is worse than the laptop you are reading this on.

Why error correction was the whole story

A qubit holds a superposition, and that state decays in microseconds from vibration, heat and stray fields. Every operation introduces error. For decades the field faced an apparent dead end: adding more qubits added more error faster than it added capability.

The theoretical answer, error correction, spreads one logical qubit across many physical ones so that errors can be detected and fixed without measuring the state directly. It requires physical error rates below a threshold, and reaching that threshold took thirty years. The 2024 results showing that a larger error-corrected patch had a lower error rate than a smaller one were the moment the curve turned the right way.

That is the difference between a science project and an engineering roadmap. The remaining work is enormous and it is engineering.

What the machines might actually be used for

Simulating quantum systems. This is the original idea and the least speculative. Molecules are quantum objects, and classical computers approximate them expensively. Catalysts, battery chemistry, nitrogen fixation and drug binding are all in this category. If quantum computing delivers commercial value first anywhere, it will be here.

Optimisation. Widely claimed, weakly supported. Classical algorithms are extremely good at practical optimisation, and most claimed quantum advantages have been matched by better classical methods once someone looked properly.

Cryptography. Shor's algorithm factors large numbers efficiently, which would break the public-key cryptography securing the internet, including the signatures behind every crypto wallet. The qubit requirements are far beyond current machines, by several orders of magnitude.

The honest timeline on the cryptography question

Breaking a 2048-bit key is estimated to need millions of physical qubits with error correction. Current systems have hundreds to low thousands of physical qubits. That is not a gap that closes next year.

It is also not a reason to relax, because of the harvest-now-decrypt-later problem: encrypted data recorded today can be stored and decrypted whenever a capable machine exists. For anything that must stay secret for a decade, the migration to post-quantum algorithms is a present-tense task, and standards for them were finalised in 2024. Most large institutions have started; most small ones have not.

For Bitcoin and similar systems, the exposure is real but manageable and well understood: signature schemes can be changed by a protocol upgrade, and the practical difficulty is coordination and dormant coins, not mathematics.

How to read a quantum announcement

Ask three questions. How many logical, error-corrected qubits, not physical ones. What is the error rate per logical operation. And was the demonstrated task chosen because it is useful, or because it is hard for classical computers and useless for anything else. Most headline results fail the third question, and saying so is not cynicism; it is how the field measures itself.

What to watch

Watch logical qubit counts, which are a small fraction of physical counts and the only number that scales to usefulness. Watch for a chemistry result that a classical method could not reach. And watch post-quantum migration in banking and government, because that is where the cost of this technology is being paid today.