MODULE 5 · LESSON 2
Free — no login requiredSign in to track progress, save quiz attempts and enrol in the full course.
Sign in to track progress / enrolUsing One Today
Quantum computers are more accessible than most people assume. Several are available over the internet, some at no cost, and running your first circuit takes an afternoon.
Cloud access
Every major platform offers hosted access. You write a circuit in a Python library, submit it, it is queued, it runs, and results return as counts of measurement outcomes across many shots.
The main software frameworks are open source and free: Qiskit, associated with IBM, and Cirq, associated with Google, are the most widely used, with several others in circulation. All are ordinary Python packages installed in the usual way.
The workflow is unlike classical programming in one important respect. You do not get an answer, you get a distribution. Run 1,024 shots and receive counts for each observed bit string. Interpreting that distribution is part of the work, and separating signal from noise in it is much of the difficulty.
Simulators do most of the work
Most quantum programming happens on classical simulators rather than real hardware, and for good reasons.
Simulators are free, immediate, and free of noise, so you can confirm an algorithm is correct before contending with hardware imperfection. They also let you inspect the internal state, which is impossible on real hardware and invaluable for learning and debugging.
The limit is the exponential wall from Module 1. Simulating around 30 qubits is comfortable on a laptop, around 40 needs a serious machine, and beyond about 50 becomes infeasible in general. That boundary is itself the clearest demonstration of why quantum computers are interesting.
The practical implication for anyone learning: you can go a long way without touching hardware, and you should, because the concepts are the same and the feedback loop is far faster.
Hybrid algorithms
Since NISQ machines cannot run deep circuits, nearly all current practical work uses hybrid approaches.
The pattern is consistent. A classical computer holds the main loop. It prepares a short parameterised quantum circuit, runs it many times, collects the measurement statistics, computes something from them, adjusts the parameters, and repeats. The quantum circuit is a subroutine inside a classical optimisation.
The best known examples are the variational quantum eigensolver, aimed at finding molecular ground state energies, and the quantum approximate optimisation algorithm, aimed at combinatorial problems.
Two honest observations about these.
They are designed around a hardware limitation. Short circuits are used because long ones do not survive decoherence, not because short circuits are theoretically preferable. Whether hybrid methods retain their importance once error correction works is genuinely uncertain.
Results have been mixed. Numerous published hybrid results have subsequently been matched or beaten by classical methods, which is the dequantisation pattern from Module 1. This is normal scientific progress and it means claims in this area deserve the classical baseline question from Module 4.
What to expect from an experiment
If your organisation runs a quantum pilot, calibrated expectations prevent disappointment and prevent overclaiming.
You will learn what your problem looks like as a quantum circuit. Genuinely valuable, and frequently the main result. Many problems turn out not to map cleanly at all, which is useful to discover early and cheaply.
You will not beat your classical solver. On current hardware, for essentially any real business problem, the classical method wins. Anyone promising otherwise should be asked the Module 4 question.
You will build capability. People who understand what these machines do and do not offer become valuable, and that understanding takes time to develop. A pilot's honest purpose is usually organisational readiness.
You should run the classical comparison. Always. It is the difference between an experiment and a marketing exercise, and it is the single most common omission in published pilot results.
Removing the mystery entirely, here is a complete first program in plain description.
You create a circuit with two qubits. You apply a Hadamard gate to the first, putting it in an equal superposition. You then apply a controlled-NOT gate with the first qubit as control and the second as target, which flips the second qubit only when the first is 1. Finally you measure both.
That is roughly five lines of Python.
What you have built is a Bell pair, the simplest entangled state from Module 2. The two qubits are now correlated: neither has a definite value, but they are guaranteed to agree.
Running it 1,000 times on a simulator gives roughly 500 results of 00 and roughly 500 of 11, and essentially none of 01 or 10. The qubits always agree, though which value they agree on varies randomly. That is entanglement, produced on your own machine in a few minutes.
Running the same circuit on real hardware gives something slightly different: perhaps 480 of 00, 490 of 11, and a scattering of 01 and 10 results that should be impossible. Those stray counts are decoherence and gate error made visible, and seeing them is the most instructive part of the exercise. The gap between the simulator's clean result and the hardware's messy one is the entire engineering problem of this field, displayed in a histogram.
Anyone comfortable with basic Python can do this in an afternoon, and it converts several abstract ideas from this course into something concrete. It is the single best next step for a learner who wants to go further.
Your organisation runs a quantum computing pilot on a real business optimisation problem. What is the most realistic expected outcome?
Shots
Click to flipRepeated runs of a circuit. Each yields one sample, so results arrive as a distribution of outcome counts rather than an answer.
Click to flip backQuantum hardware is accessible over the cloud through free open source Python frameworks, and results arrive as distributions of outcome counts across many shots rather than as answers. Most learning happens on classical simulators, which are free, noiseless and inspectable up to roughly 30 to 40 qubits, and that ceiling is itself the exponential wall. Current practical work is dominated by hybrid algorithms in which a classical loop calls short circuits, a design forced by noise rather than chosen. Expect a pilot to yield insight and capability, not a win over your classical solver, and always run the classical comparison.