Magnetohydrodynamics. Two dimensions. Zero shortcuts.
A 128×128 ideal-MHD solver, running continuously. Orszag-Tang initial conditions on a 2π periodic domain. It drives the public entropy beacon — and its conservation numbers are not asserted in copy, they are read off the signed frame header below, where you can check the signature yourself.
Lax-Friedrichs · Δm/m and ΔE/E measured from signed frame headers
Ideal MHD in conservation form. 6 field components per cell (ρ, vx, vy, p, Bx, By). Lax-Friedrichs fluxes on a 128² periodic grid. First-order in space — diffusive at shocks, exact where it counts.
The canonical 2D MHD test. Smooth initial conditions evolve into turbulent small-scale structure with sharp shock fronts. Mass and energy drift are measured against their step-zero anchors every step and shipped in each frame's header — the readouts on the viewport are parsed from those bytes in your browser, never typed by us.
Density (ρ) sampled from /entropy/frame/current — the solver grid,
min-max normalized and color-mapped in your GPU. A new keyframe lands about
every ~2 s;
the display cross-dissolves between keyframes — a video fade, not fabricated physics.
Exactly what is signed: an independent witness publishes an Ed25519 signature over the
SHA-256 of the latest frame at /entropy/frame/latest.attestation.json, refreshed
about twice a minute. The ⌁ buttons re-fetch the frame, re-derive its hash in your
browser, and check that signature — if the frame has rotated past the signed one,
the proof says so instead of pretending.
Plasma is chaotic. Small perturbations blow up fast. That makes it a good random-source — the output of a correctly-running MHD step is unpredictable even to us, which is the property we need for an honest entropy beacon. And it's beautiful to look at.
There's a serious version of this argument involving fusion reactor control, but we'll let the beacon make the case.