Prove Einstein wrong. Again.

Holders set a real quantum experiment. Each time it beats Einstein, it mints random numbers nobody can predict, not even us, and posts them on Solana for any app to use.

Your move: set the left analyzer.

0 of 16 choices keys 1 2 or ← →

Practice runs are simulated in your browser. : signed-in holders’ choices run on the machine every minute.

the mould: the ideal bell, cos 2θ

	b = 22.5°	b′ = 67.5°
a = 0°	+0.71predicted	−0.71predicted
a′ = 45°	+0.71predicted	+0.71predicted

S = E(a,b) − E(a,b′) + E(a′,b) + E(a′,b′)

Fig. 1. The apparatus. Each pair leaves the crystal entangled; each analyzer answers +1 or −1. The bell above is the correlation curve E(θ) = cos 2θ over one period, cast from measurements; the needle shows S. Your choices turn the left analyzer, here or on the dials above.

Randomness Einstein can’t rig

When a batch beats Einstein’s limit by enough, the violation itself proves its outcomes were unpredictable. BELL distils them into certified random bits and posts each batch as a pulse on Solana, chained to the one before. Raffles, game loot, airdrop picks and launch orders read it; each read is paid in $BELL, and those payments are burned.

Fig. 2. The beacon. Values ride the lines: the latest real pulse, or your practice run once you make one. live runs now, later waits for the hardware key or the coin.

Beacon tape

Live from the BELL chain; every pulse links to a full recomputation. The CURBy row is the latest pulse of the University of Colorado beacon. Read the beacon from your app.

How it’s certified. If the outcomes were predictable, S couldn’t exceed 2. Past it, the violation bounds how well anyone could have guessed each outcome: at least 1 − log₂(1 + √(2 − S²/4)) bits of min-entropy per pair, the standard device-independent bound, taken at S − 3σ to be safe. A Toeplitz hash seeded by the previous pulse turns 4,096 raw outcomes into that many uniform bits, minus 64 for a 2⁻³² margin.

What you have to trust. On one chip the two qubits sit millimetres apart, so certification assumes they don’t signal each other. Every pulse also mixes in the latest pulse of CURBy, the University of Colorado Randomness Beacon built with NIST on a loophole-free Bell test (Kavuri et al., Nature, 2025). Its quantum chain last published on 22 August 2025; its public chain still pulses every minute.

Not just us. In March 2025, JPMorgan Chase and Quantinuum reported certified randomness from a 56-qubit trapped-ion computer in Nature, with a different protocol. Same idea: randomness you can check.

BELL publishes randomness. It doesn’t run raffles, games or lotteries.

The foundry

Your run fills the mould. The shape, S and every flaw come from your measurements, so they can’t be edited. The inscription and the face are yours.

Fig. 3. The dashed mould is the ideal bell, cos 2θ. Metal only goes where your measurements put it.

Set by the physics

S: —
Past Einstein’s limit: —
Visibility: —
Rarity: —
Flaws in the casting: —

Yours

Inscription 28/40

Face

The result card

Every experiment pairs two holders who never meet. Each keeps a card with a bell cast from that experiment’s own measurements: a clean run rings true, a noisy one comes out flat or cracked. Plus both holders’ choices and the S the machine returned. Cards are minted by burning $BELL.

The latest experiments, live from the API.

Einstein’s last way out

In 1947 Einstein wrote to Max Born that quantum mechanics seemed to need something he could not accept. He gave it a name.

„ohne spukhafte Fernwirkungen“

“free from spooky actions at a distance”

Einstein to Born, 3 March 1947

In 1964 John Bell turned the argument into a test, and in 1969 Clauser, Horne, Shimony and Holt gave it a number. If the world works the way Einstein hoped, a quantity called S can never go above 2. Quantum mechanics says it can reach 2√2, about 2.83.

Experiments keep landing above 2. The 2022 Nobel Prize in Physics went to Alain Aspect, John Clauser and Anton Zeilinger for this work.

One way out remained: maybe the settings were never really free. So on 30 November 2016, about 100,000 people played an online game whose choices set 13 experiments in 12 labs on five continents. BELL runs it again, on-chain. Your choices set the experiment. The machine answers.

Fig. 5. Published values of S, with their reported uncertainties, in the order they were measured. Every one sits past Einstein’s limit. Sources in the footer.

1935Einstein, Podolsky and Rosen ask whether quantum mechanics is complete.
1964Bell turns the question into a test.
1982In Orsay, Aspect’s team measures S = 2.697.
2022The Nobel Prize goes to Aspect, Clauser and Zeilinger.
2026BELL holders, running S —

Fig. 6. Ninety-one years of one question. The last point is yours.

The machine, now

The quantum meta talks about IBM Quantum machines. BELL gives one the test the whole field grew out of: two qubits entangled, turned to the holders’ angles, read out. Sixteen choices a side at 256 shots each, plus a 16-angle sweep that casts the card’s bell: 8,192 shots per experiment.

Fig. 7. S for every experiment so far, one dot each with its error bar, and the running weighted mean in orange.

Global S: —live
Experiments: 00 pairslive
Certified bits: 00 pulseslive
Halloween night: —to 31 Oct, 00:00 UTClive
Next pairing: —live
IBM Quantum status: checking…live

Fig. 8. The circuit. A Bell pair, then each qubit is turned by twice its holder’s angle before readout.

On Halloween night, every experiment is printed twice: its card carries a second, ghost impression of its number.

Ledger

Latest experiments, newest first.

Exp.	UTC	Holder A	Holder B	Hardware	S	Past the limit	Result

Table 1. Every experiment, newest first. Each finished row opens its card; the API serves the raw outcomes, the seed or IBM job, and a verification of every pulse.