an interactive paper · parallax engineering · 2026
Twelve results from quantum mechanics that sound like mysticism — superposition, entanglement, nonlocality, interference, the quantum eraser, Feynman's sum over histories, tunneling, uncertainty — demonstrated not with metaphor but with live computation. Every figure on this page is being calculated in your browser right now: a Schrödinger equation is being integrated, Bell inequality trials are being tallied, wavefunctions are collapsing. Nothing is a video. Everything is math.
The claim of this document is simple: the mystical-sounding parts of quantum mechanics are its most rigorously proven parts — and the same equations we prove them with are the ones we now build machines out of. Scroll down and check the numbers. And when the proofs run out, we will show you exactly where.
▽ begin the argumentEight things a particle can do that a marble cannot.
A qubit is the simplest quantum object: a system with two answers, $|0\rangle$ and $|1\rangle$. Its full state is a point on a sphere — the Bloch sphere —
The poles are the definite answers. Everywhere else on the sphere is a superposition — a state with genuine amplitude for both outcomes at once. The mystical claim "it's both until observed" is literally the geometry: almost every state on the sphere is neither pole.
Drive the state yourself. Apply gates — real unitary rotations — and then measure. Measurement samples $P(0)=\cos^2\frac{\theta}{2}$ and snaps the state to a pole. Run it repeatedly from the equator and watch the 50/50 statistics accumulate: the superposition was real, and the collapse is real too.
Superposition was one qubit. Take two. Most two-qubit states factor: qubit A is in some state, qubit B is in some state, and the pair is just the two of them side by side. But some states do not factor at all — and those are where the trouble starts.
Build one. Apply a Hadamard to qubit 0, then a CNOT (qubit 0 controls qubit 1). You get a Bell state:
Neither qubit now has a state of its own — ask for qubit 0 alone and the honest answer is "50/50, no more to say." The information isn't in either particle; it lives entirely in the correlation. Measure one and the other's outcome is fixed instantly, however far apart they are.
Watch the concurrence climb from 0 to 1 as you build it. This is not a metaphor for the next section — it is its raw material. The correlations we are about to march through Bell's theorem are made right here, on this canvas.
This is the strongest result on the page — a 2022 Nobel Prize, and the closest physics comes to proving something spooky. Take the entangled pair you just built (here in its singlet form $\;|\Psi^-\rangle = \tfrac{1}{\sqrt2}\big(|01\rangle - |10\rangle\big)$), separate the two qubits, and measure each along chosen axes.
Bell's theorem (1964): if each particle carried a pre-existing answer — any local hidden plan whatsoever, however clever — then the CHSH statistic obeys
Quantum mechanics predicts $E(a,b) = -\cos(a-b)$, which at the optimal angles reaches $S = 2\sqrt{2} \approx 2.828$. Both can't be right.
Below, two experiments run head-to-head: a genuine quantum singlet sampler, and the best possible local hidden-variable model, each measured at the same four angle pairs. Run the trials. The classical model saturates at 2 and stops. The quantum one walks straight through the wall — with error bars.
Fire electrons at two slits, one at a time, so no electron can possibly interfere with another. Each arrives as a single dot. And yet the dots accumulate into interference fringes — the signature of a wave that went through both slits:
That cross-term is the whole mystery. If the electron had secretly picked one slit, probabilities would just add — no fringes. The fringes are a measurement of the fact that it didn't pick.
Now the twist the mystics love, correctly: put a which-path detector at the slits. Learn which slit each electron used — don't block anything, just know — and the cross-term dies. The pattern becomes two boring humps. Information, not force, changes the physics.
Proposition IV said that knowing the path destroys the fringes. Here is the twist that sounds like time travel and isn't. Tag each electron with a marker qubit that records which slit it took. The fringes vanish, as promised — the path information now exists, whether or not anyone has looked.
But the marker is itself quantum. Instead of reading it in the which-path basis, read it in a complementary basis — a measurement deliberately blind to which slit, which erases the path information rather than revealing it. Now sort the electrons by that erased marker's two outcomes.
Within each subset the fringes return — but shifted: one subset lands in fringe, the other in anti-fringe, offset by $\pi$. Add the two subsets back together and they wash out again, which is exactly why this can never send a signal. The interference was always there, hidden in a correlation; erasing the marker is what lets you sort it back into view.
You can even choose to erase after every electron has already hit the screen. Nothing in the past changes. What changes is which correlations you are permitted to sort the same dots by.
Where do the "two paths" of the double slit even come from? Feynman's answer is outrageous and correct: from all of them. A particle travelling from A to B takes every conceivable path at once, and each path contributes a unit-length amplitude that spins by the classical action $S$ along it:
Most paths have wildly different actions, so their phasors point every which way and cancel in pairs. Only near the path where the action is stationary, $\delta S = 0$, do neighboring paths share nearly the same phase and add up. That surviving bundle is the classical trajectory.
So classical mechanics isn't a separate law — it's what's left standing after everything else interferes itself away. Turn $\hbar$ down with the slider and watch the surviving bundle narrow to a single line: the classical limit is a limit of this sum, not an escape from it.
This figure is a real numerical solution of the time-dependent Schrödinger equation,
computed ~60 times a second by split-step Fourier (the kinetic term is applied in momentum space via FFT — the same algorithm used in research codes).
A wave packet with energy $E$ hits a barrier of height $V_0 > E$. Classically the transmission is exactly zero: the particle does not have the energy, full stop. Quantum mechanically, the wavefunction decays inside the barrier instead of stopping, and leaks through with probability $\;T \approx e^{-2\kappa L},\; \kappa = \sqrt{2m(V_0-E)}/\hbar.$
Watch the transmitted probability tick upward on the right of the wall. That number is why the sun shines (proton fusion is classically forbidden), why flash memory works, and why alpha decay happens. "Walking through walls" isn't a metaphor here — it's a boundary condition.
The uncertainty principle is routinely mis-told as "measuring disturbs." The truth is stranger and cleaner: position and momentum are Fourier transforms of each other. They are two readings of one object, and a function cannot be narrow in both readings:
This is a theorem about waves, older than quantum mechanics — the same reason a short audio click has no definite pitch. The particle doesn't hide its momentum; a particle with a sharp position does not possess a sharp momentum, any more than a chord possesses a single note.
Squeeze the position distribution with the slider. The momentum distribution — computed live as the actual Fourier transform of $\psi(x)$ — flares out in exact compensation. The product $\sigma_x\sigma_p$ is printed below; for a Gaussian it pins the theoretical floor $\hbar/2$ precisely. And because the copy above isn't just an analogy — hear it: play $\psi(x)$ as a sound and the same tradeoff is a click versus a pure tone.
The same equations, turned into hardware you can hold, a chip you can buy — and quantum information you can move.
You already know this circuit. An inductor and a capacitor trade energy back and forth at $\omega = 1/\sqrt{LC}$. Charge on the capacitor plays the role of position; flux through the inductor plays the role of momentum. It is, term for term, a harmonic oscillator:
Quantize it and the energy stops being continuous. It comes in evenly spaced rungs — including a $\tfrac12\hbar\omega$ you can never remove, the zero-point energy of the vacuum itself. Cool a real superconducting LC circuit to a few millikelvin and these rungs are laboratory facts.
One catch: because the rungs are evenly spaced, a pulse tuned to drive $0\to1$ also drives $1\to2$ — you can't isolate two levels to use as a bit. Add a Josephson junction, a nonlinear inductor, and the ladder becomes anharmonic: the $0\to1$ gap no longer equals the $1\to2$ gap, so you can address just the bottom two rungs. That engineered two-level system is the transmon — the qubit inside most quantum computers being built today.
A single atom has sharp, isolated energy levels. Bring a second atom close and each level splits in two — an electron can now hop between the wells, and the exclusion principle forbids two identical states, so they repel into a bonding and an antibonding level. Add more atoms and each level splits again, and again.
Push it to a crystal — $10^{23}$ atoms — and the splittings merge into continuous bands separated by forbidden gaps. A tight-binding chain gives the dispersion exactly:
a band of width $4t$ set entirely by how strongly neighbors couple. Whether the highest occupied band is full or only partly full — and how wide the gap above it is — decides everything: metal, insulator, or the tunable in-between, the semiconductor.
That gap is the thing you open and close when you switch a transistor. The most consequential technology of the last century is this figure, filled with electrons up to a line. Drag the atom count from one to many and watch levels become bands.
Here is everything on this page at once. Alice has a qubit in an unknown state $|\psi\rangle$ — she cannot measure it (that would collapse it) and the no-cloning theorem forbids her from copying it. Yet she can send it to Bob, intact, without either of them ever learning what it is.
The trick spends one unit of the resource you built in Proposition II. Alice and Bob pre-share an entangled Bell pair. Alice measures her unknown qubit together with her half of the pair, in the Bell basis — getting two random classical bits and, in the act, destroying $|\psi\rangle$ at her end:
The instant she measures, Bob's qubit is already a version of $|\psi\rangle$ — but scrambled by a Pauli gate he can't guess. She sends him the two bits over an ordinary classical channel; he applies the matching correction and his qubit becomes $|\psi\rangle$ exactly.
Watch the three "no"s hold simultaneously: no cloning (Alice's copy is gone the moment Bob's appears), no measurement of $|\psi\rangle$ (nobody ever learns $\theta,\varphi$), and no faster-than-light (until the classical bits arrive, Bob's qubit is useless — the fidelity readout sits below 1). Entanglement alone moves nothing; it takes two humble classical bits to finish the job.
Everything above was proven. This is the one thing that isn't.
Every figure so far computed a definite answer. This one computes the exact place where physics stops being able to. The Schrödinger equation is deterministic and never collapses — it only spreads amplitude around. Yet every measurement returns a single outcome. The gap between those two sentences is the measurement problem, and after a century it is still unsolved.
Couple a qubit to an environment and watch decoherence. Write the state as a density matrix,
The diagonal entries are the populations (the odds of 0 vs 1); the off-diagonal entries are the coherences — the genuine "and" of superposition. An environment that "watches" in the 0/1 basis makes the coherences decay, $\;\rho_{01}(t) = \rho_{01}(0)\,e^{-t/T_2}$, while the populations stay put. The Bloch vector shrinks from the surface toward the centre; purity $\mathrm{Tr}(\rho^2)$ falls from $1$ to $\tfrac12$. A pure superposition becomes a classical-looking mixture. This is real, and it is being integrated live in the figure.
Now watch what it doesn't do. Decoherence turns "both" into "either/or" — it explains why you never see a blurred pointer or a half-dead cat. It does not pick an outcome. Press measure: the odds are exact, but which result becomes real, and by what rule, is written in no equation we have. Many-Worlds says every branch happens and you ride one. Objective-collapse models say a real physical process — perhaps gravity — selects one. Copenhagen declines to say. A hundred years in, we cannot tell them apart.
The universe hands you the probabilities and keeps the mechanism. That is the edge of the map — and the honest end of this argument.
The instrument this whole paper was building — now pointed at the frontier.
You've walked the whole range now — from ironclad proof to the genuinely unknown. That range is the instrument. A claim about the quantum frontier isn't true or false because it sounds spiritual or sounds skeptical; it earns its place by where it lands on that scale. So point the scale outward.
Parallax is the whole method, and it's the reason for the name: you cannot judge a distance with one eye. The believer ("entanglement means we are all one") and the debunker ("it's all nonsense") are each squinting through a single vantage, and both miss the real depth. Triangulate from both and a claim resolves into three layers: what's actually real, what's genuinely open, and what's overclaimed.
The blade that separates them is falsifiability. A claim worth your time makes a prediction that could kill it — and hasn't been killed. One that forbids no observation isn't a hypothesis; it's a vibe. We don't debunk. We triangulate: honor the mood, correct the mechanism.
Start with the most seductive claim — that entangled particles let minds, or bodies, touch across space. Its kernel is real: the correlations are nonlocal (Proposition III). So watch Alice spend a genuine quantum resource trying to send Bob a single bit with it. Bob's needle never leaves 50%. The spookiest thing in physics has an information channel capacity of exactly zero.
Six claims from the quantum fringe, read by parallax. Each links to the proposition that settles it.
Signal. Measurement is genuinely unresolved — nothing in the equations selects a single outcome.
Noise. Consciousness appears nowhere in the math. A stray photon, a detector, the air — any interaction decoheres the state. Rocks "observe" each other.
Kill test. Do unwatched detectors still log definite results? They do — refuting the consciousness requirement.
↳ Prop XII · the measurement problemSignal. Erasing which-path information really does restore an interference pattern — even "after" detection.
Noise. Nothing in the past changes. You only choose which already-recorded data you're allowed to sort by. No effect precedes its cause.
Kill test. Read the erased outcome before the choice is made? Impossible — the subsets are defined by the choice.
↳ Prop V · the quantum eraserSignal. The correlations are perfectly real and perfectly nonlocal: measure here, and the far outcome is fixed at once.
Noise. The no-signaling theorem forbids using it. Bob's local data is 50/50 noise until Alice mails her results classically. Capacity: zero.
Kill test. Send one bit through entanglement alone (Fig. 13). Decades of trying: it can't be done.
↳ Prop III & the demo aboveSignal. Zero-point energy is real — the ground state still holds ½ħω you can never remove, as the LC ladder shows.
Noise. "Ground state" means lowest — there's no lower level to fall to, so no work can be extracted. A well with no downhill.
Kill test. Build the perpetual machine and measure its output. Net energy budget: exactly zero.
↳ Prop IX · the quantized oscillatorSignal. Not grift — a serious, if fringe, Penrose–Hameroff hypothesis with a real falsifiable core.
Noise. The brain is warm, wet, and noisy — exactly where decoherence is fastest. Estimates put coherence times ~10⁻¹³ s against neural timescales of ~10⁻³ s.
Kill test. Measure coherence lifetimes in microtubules at 310 K. So far the gap is ~10 orders of magnitude — but the question is live.
↳ Prop XII · decoherence timescalesSignal. The only real ingredient is the word. Quantum mechanics does govern the molecules — as it governs a brick.
Noise. No mechanism, no energy field, no dose, no measurable claim. Any coherence at body temperature dies in femtoseconds.
Kill test. There isn't one — the claim forbids no observation. That's the tell: it's marketing wearing a lab coat.
↳ Prop VIII · what "quantum" actually constrainsNotice the pattern: every real kernel makes a prediction someone went and measured. Every overreach makes none. That single difference — not who it offends — is the whole method. The mystics were wrong about the mechanism and right about the mood.