Quantum computing

This page is for readers who want to know whether the advice rests on real understanding. It does. Here is a plain account of where the technology stands and where I work in it.

The model, briefly

A quantum computer manipulates qubits through gates, exploiting superposition and entanglement to evaluate structured problems in ways a classical machine cannot efficiently reproduce. The headline algorithms set realistic expectations:

• Shor's algorithm factors integers and solves discrete logarithms efficiently. This is what breaks RSA and elliptic-curve cryptography. It needs a large, fault-tolerant machine: many logical qubits built from many more physical qubits. That is the timeline driving post-quantum migration.
• Grover's algorithm gives a quadratic, not exponential, speedup for unstructured search. Practically, it halves the effective strength of symmetric keys, which is why the answer is to move from AES-128 to AES-256 rather than to panic.
• Variational methods (VQE, QAOA) are the NISQ-era workhorses: hybrid quantum-classical loops aimed at chemistry, materials and combinatorial optimisation. Promising, genuinely hard to make beat classical methods today, and where much of the honest near-term research sits.

Where we really are: the NISQ era

Current devices are noisy and shallow. Decoherence limits how many gates you can run before the signal is lost. Two distinct strategies matter, and they are often confused:

• Error mitigation (zero-noise extrapolation, probabilistic error cancellation, and similar) cleans up results in post-processing. It helps now but does not scale indefinitely.
• Error correction encodes one logical qubit across many physical qubits so errors can be detected and fixed below a threshold. This is the path to fault tolerance and to running Shor at scale. Early logical-qubit demonstrations are encouraging; useful fault-tolerant machines are still a multi-year effort.

"Quantum advantage" has been shown on contrived sampling tasks chosen to be hard for classical computers. It has not been shown on a practically useful, commercially relevant problem. I say so plainly, because over-promising is how quantum projects lose their budget.

How I work in it, hands-on

I build and reason about circuits in Qiskit, prototype variational and search-style algorithms, and read the primary literature rather than the press releases. The point is not to run production quantum workloads, which almost no one should yet. It is to give you advice grounded in what the machines and algorithms can and cannot do, so an opportunity assessment is realistic rather than aspirational.