Choose an experiment. Ask your question, start the simulation and explore what the result means. Live calculations run on our i9, with no account required.
Six wheels. 1,000,000 possible codes. Set a test code, scramble the wheels and let our i9 search for it.
Your test code: 314159
Drag the numbers up or down with your mouse or finger, or use the arrow buttons. On a keyboard, use ↑, ↓ or the number keys.
Set your chosen code on the open lock. Closing it saves this test code as the matching rule.
The compute worker is currently unavailable. You can still set up the lock.
Real quantum simulation. Your code. Our computer. Our i9 runs a simulation of Grover's quantum program with 20 search qubits on all ten cores. Once the result is confirmed, the wheels settle on the matching code and your virtual lock opens.
The lock is virtual; the calculation is real. Our ordinary computer runs a mathematical simulation of the quantum program. Please use a made-up test code: it is sent to our server for the calculation.
You supplied the test code. It defines the matching rule for our quantum program. Grover's algorithm then increases the probability of measuring the matching result. An unknown real PIN would first require a suitable checking rule; this learning lock is not a way to open real locks.
The lock has 1,000,000 valid codes, from 000000 to 999999. Twenty search qubits represent 1,048,576 bit strings; the remaining 48,576 are not valid codes. Our specialised C++ simulator computes all 1,048,576 possibilities using 804 Grover rounds, with no extra helper qubit. The most frequent measured valid code must satisfy the matching rule, or the lock stays closed.
The ideal success probability per measurement is about 99.99998%, not a guaranteed 100%. The full double-precision state vector occupies 16 MiB; the complete program requires additional memory. The calculation uses all ten cores of the i9 with AVX2 instructions. The result is bit for bit the same as with a single core.
More digits mainly mean more computing work: in each of the 804 rounds, the simulation recomputes all 1,048,576 possibilities. That takes about a tenth of a second. An ordinary program that checks the codes one after another against the same rule would finish sooner here. The simulation shows how the quantum program works. It is not a speed record and not a quantum advantage.
The wheels do not show quantum codes being tested one by one. Their fast movement fills the wait for results; slow motion then highlights the match. Both are visual design, not a measure of compute speed. With reduced motion, the result appears directly.
Four small quantum programs. Predict the outcome, then let our i9 system simulate the circuit and draw 2,048 measurement samples. Real computation, no sign-up. The treasure hunt supports up to 12 search qubits plus a helper qubit.
The qubit starts in 0. The H operation creates an equal superposition of 0 and 1. Then the simulator samples a measurement. It repeats this 2,048 times, with a freshly prepared state each time.
Each sample is random; its probability is not. Equal probabilities give roughly equal counts, not necessarily exactly equal counts.
Apply H twice without measuring in between. Two ordinary coin flips would remain random. For a qubit, something different happens.
The amplitudes for 1 cancel and those for 0 reinforce. This is interference, the idea illustrated by our maze. Ordinary random coin flips cannot reproduce it.
Two qubits: H creates a superposition on the first, then a controlled flip entangles them. Measure both. Which pairs appear?
This pair is a Bell state. Measuring either qubit alone is random; in this basis the two results always match. It cannot send a message: the correlation appears when records are compared.
Four hiding places, one treasure: you choose its location. The program marks the matching state and amplifies it. Will one measurement find it?
Marking flips the target's sign; amplification cancels the other possibilities. With four hiding places, one ideal round is sufficient. Below, try the same idea with a larger search space.
From 6 to 12 search qubits on the same i9: 4,096 instead of 64 hiding places, a search space 64 times larger. Start small or choose the larger size. The 12-search-qubit circuit uses an additional helper qubit. Grover's quantum program amplifies the matching outcome in 6 rounds for this search size. Our ordinary computer simulates the circuit and then counts 2,048 measurement samples.
6 search qubits · 64 possible search results · 6 Grover rounds · 2,048 samples
2 to the power of 6 is 64: one more search qubit doubles the number of possible outcomes. For 256 or more hiding places, a reusable helper qubit lets Qiskit decompose large operations into fewer small gates. This is a more efficient construction of the same Grover algorithm, not a new search method.
For 64 hiding places we use 6 rounds. The rule of thumb is roughly π/4 times √N. This compares calls to a matching oracle, not seconds: a Grover round contains many gates. Classical search can be much faster than this simulation on our ordinary computer.
You provide the target so we can check a known solution. The program contains a matching oracle. This does not prove faster searching of unknown real-world data. For 4,096 possibilities, we separately compared the full simulated state with the mathematical prediction.