Each experiment is a different test that no classical system can pass. The draw works the same whichever runs; what changes is the physics measured, the proof published that day and the look of the winner's certificate. All three already ran on real IBM hardware.
Results marked real were measured on IBM quantum hardware on 10 Oct 2026. Vote with your wallet on each card: a free signature, no transaction. Voting for a draw closes at 12:00 UTC; the final count at 17:50 UTC is weighted by the tokens each wallet holds, and with no votes the three experiments take turns.
Two neighbouring qubits on the processor's lattice. A Hadamard puts the first in superposition, a CNOT entangles the second with it. Each qubit is then measured at an angle picked at random, thousands of times. The joint counts give the Bell value S, and the raw outcomes, sealed with a drand round, become the day's number on Solana.
H on q₀, then CNOT q₀→q₁: the pair is now in the Bell state |Φ⁺⟩.
q₀ is read at 0° or 90°, q₁ at 45° or −45° on the Bloch sphere: the CHSH settings.
Resonators read each qubit as 0 or 1. Thousands of shots build the joint counts.
Counts give S. The outcomes are hashed with a future drand round and written to the Solana oracle.
Stylised processor view. Real runs use IBM's native gates (the CNOT is compiled from them) and the backend available that day, for example a Heron-class chip.
For two qubits in |Φ⁺⟩ measured at angles θ₀ and θ₁, the correlation is cos(θ₀ − θ₁). Any classical model stays on or inside the straight line. At the CHSH angles the gap adds up to S = 2√2 ≈ 2.83 in theory; real hardware lands lower because of noise, and anything above 2 shows quantum entanglement.
Bars: live illustrative counts with about 6% hardware noise (sample S 2.64). The value on the right and the points on the curve were measured on IBM hardware.