Amplify. Twist. Repeat.
That is all a quantum gate does. A complex amplitude has a size and an angle, and every gate stretches one and rotates the other, which is where this site gets its name. Chain enough of them and the wrong answers cancel while the right one survives. Here is where I work through how, with interactive simulations you can push around, plain-language explanations, and honest notes on where quantum computing might eventually matter versus where the claims currently outrun the evidence.
Three problem families worth watching
Quantum advantage is not general-purpose. It shows up in narrow places with the right structure: rugged optimisation landscapes, and systems that are themselves quantum. These three are where I think the case is strongest, and they are what I am reading toward. To be clear: this is where my interest points, not a description of work I am currently doing.
Combinatorial optimisation
Scheduling, routing and resource allocation are NP-hard and turn up everywhere in industry. Ising and QUBO encodings turn them into something a quantum device can attack, though the classical solvers they would have to beat are very good.
Materials discovery
Electronic structure is a natively quantum problem. Variational eigensolvers estimate ground-state energies of lattice and molecular Hamiltonians that scale badly classically.
Molecular simulation
Binding affinity and conformational search for drug candidates. Today this means small active-site models and hybrid quantum-classical pipelines, not whole proteins.
What a quantum computer might eventually be used for
A quantum computer is not a faster computer. It is a different one, useful only where a problem has structure that interference can exploit. Below is the honest list of candidate applications, each with a horizon attached: near term means devices that exist today already do this, early advantage means the case is credible but unproven on real instances, and fault tolerant means it waits on error-corrected machines that nobody has built yet.
Chemistry and catalysis
Nitrogen fixation, the reaction behind most of the world's fertiliser, still runs on a century-old process that burns roughly one percent of global energy. Nature does it at room temperature with an enzyme nobody can simulate accurately. Simulating that active site is the classic case for a quantum computer, because the problem is itself quantum.
Batteries and energy materials
Cathode chemistry, solid electrolytes and candidate superconductors all hinge on strongly correlated electrons, which is exactly where classical approximations break down. A reliable ground-state energy for a few dozen orbitals would change how these materials get screened.
Drug discovery
Not designing whole molecules, despite what the press releases say. The realistic target is the small quantum-mechanical core of a binding site, treated accurately and embedded in an otherwise classical pipeline. Read the honest status of this one.
Logistics and scheduling
Routing fleets, sequencing factories, allocating scarce resources. These map cleanly onto Ising and QUBO landscapes, which is what makes them attractive. It is also the area where classical solvers are strongest, so this is the application most likely to be oversold.
Finance and risk
Derivative pricing and value-at-risk are Monte Carlo problems, and amplitude estimation offers a provable quadratic speedup over sampling. Real, but quadratic: the crossover against a well-optimised classical simulation sits far past current hardware.
Breaking today's encryption
Shor's algorithm factors integers in polynomial time, which retires RSA and elliptic curve cryptography on the day a large enough error-corrected machine exists. The consequence is already here: encrypted traffic captured now can be stored and decrypted later, which is why post-quantum standards are being rolled out today.
Machine learning
Quantum kernel methods and variational classifiers embed data into quantum states that are awkward to reproduce classically. Interesting mathematically, and the weakest of these cases in practice: most demonstrated advantages are on datasets constructed to make the quantum method win.
Sensing and metrology
The quiet success story. Entangled and squeezed states already beat the classical noise limit in atomic clocks, magnetometers and gravimeters, and LIGO uses squeezed light to hear fainter black hole mergers. This is quantum technology that works now, though it is sensing rather than computing.
Networking and key distribution
Measurement destroys the state it reads, so an eavesdropper cannot copy a quantum channel without leaving evidence. That turns a limitation of the physics into a security guarantee, and it is deployed on real fibre links today, even if the practical case against good classical cryptography is still debated.
Read that list with the horizons in mind. Two of these nine are shipping products, and the rest are bets on hardware that does not exist yet. Anyone selling you the middle seven as available today is selling something.
Quantum computing, visually
Eight short lessons, starting with the only linear algebra you need and running through qubits, superposition, phase, interference, entanglement, measurement and a real algorithm. Every concept has a live simulation you can drag, not a static diagram.
Open the course → Sandbox · PuzzlesThe quantum playground
Build circuits gate by gate and watch the state respond in real time. Make a Bell pair, break interference, cheat at a coin flip. Nine puzzles, from "make a superposition" to "build a GHZ state".
Start playing → Reading listBooks worth your time
A short, opinionated shelf, sorted by where you actually are, not by title. Each entry says who it is for, so you don't end up in a graduate text on day one and conclude the field isn't for you.
See the shelf →Nobody has won yet
Classical computing settled on the silicon transistor and stopped arguing. Quantum computing has six serious candidates for what a qubit should physically be, each with a different answer to what the hard problem is, and each with real companies and real machines behind it.
Key papers
Founder
Amplitwist is one person learning in public, not a company. Saying so plainly seems better than hiding behind a corporate "we".
Where the field actually is
Today's quantum hardware is noisy and small. For nearly every industrial problem, a good classical algorithm still wins, and saying otherwise does the field no favours. What is worth doing now is finding which problem structures might eventually invert that, building the mappings, and measuring honestly against the best classical baseline available. Anything on this site that sounds more confident than that is a mistake. Tell me.