Interactive experiments
Small interactive puzzles that turn quantum-computing ideas into something you can play with.
A Sudoku encoded as an energy-minimization problem. Generate a puzzle, then find the ground state of the same reduced QUBO objective used by the quantum solver.
Ready. Generate a puzzle or solve this one.
Browser demo: the QUBO ground state is minimized exactly so this static GitHub Pages site stays fast and dependency-free. The PennyLane/QAOA implementation uses the same objective.
A treasure is hidden in one of 16 basis states. Start from an equal superposition, amplify the marked state with Grover iterations, then measure the 4-qubit register.
Bar height shows simulator probability. These amplitudes are not directly observable on real hardware.
A new oracle is ready. The treasure is equally likely to be in any of 16 states.
Genuine 4-qubit statevector simulation: each step phase-flips the oracle-marked basis state, then applies inversion about the mean. Measurement is sampled from the resulting |amplitude|² distribution.
Put a classical random walker and a discrete-time quantum walker in the same maze. The classical agent follows one random trajectory; the quantum walk evolves coherent amplitudes over many positions, where paths can reinforce or cancel through interference.
Genuine coined quantum-walk simulation. Each open position carries four directional coin amplitudes. A step applies the 4-state Grover coin G = 2|s⟩⟨s| − I, then a flip-flop unitary shift along open edges; blocked directions act as self-loops. The heatmap exposes simulator probabilities for learning—on hardware, those probabilities would be estimated from repeated measurements.
Color a randomly generated map with only three colors so that neighboring regions never match. Or let a 12-qubit QAOA statevector search the graph-coloring QUBO for a zero-energy assignment.
New map ready. Try coloring it yourself, or run QAOA.
Genuine browser-side QAOA simulation. The graph-coloring cost is emitted as an explicit QUBO with exactly-one penalties per region and same-color penalties per shared border. Two adjacent regions are symmetry-fixed internally, reducing 18 one-hot variables to 12 qubits; the displayed solution comes from the best state actually sampled from the optimized QAOA statevector. Exact enumeration is used only as a reference.