// spacex visualization

spacex

SpaceX visualization tool. This Hopfield associative memory model is the same physics used in flight simulation, real-time 3D telemetry, and neural decoding at SpaceX and Neuralink. Imprint a pattern. Add noise. Recall. The computer should drive. Built as the minimal impressive device. Hire me to build the real production version for your mission data or Neuralink UI. Costs more than $3500. DM for quote. https://phronesis.world/work/build

imprint "thought":
temperature0.00

memories 0 · energy 0 · recalled — · empty

imprint a few · click + drag to draw your own, then break it and recall · red = north, blue = south · raise temperature to melt the memory

// the science under it

  • memory as magnetism. Treat neurons as spins that point up or down and pull their neighbors into line — the same math that describes a magnet. (Ising 1925; Little 1974)
  • imprinting is Hebbian. Storing a pattern means strengthening the link between magnets that agree in it: Wij += si sj / N. Each stored pattern becomes a low-energy valley. (Hopfield 1982, PNAS)
  • recall is rolling downhill. Flip each magnet to match the field around it and the energy E = −½ Σ Wij si sj never rises — so a broken cue slides into the nearest stored valley. That's associative memory: the whole from a fragment.
  • it can hold about 0.138 N. Past roughly that many patterns the valleys blur into spurious states and recall breaks down. (Amit, Gutfreund & Sompolinsky 1985)
  • temperature is noise. Raise it and the magnets flip at random (Glauber dynamics) — warm enough and the memory melts. The physics behind this won the 2024 Nobel Prize in Physics (Hopfield & Hinton).

// what this is not

  • neurons are not literally magnets. This is an abstraction — it shows how alignment can store and recall a pattern, not how a brain actually works.
  • this is also not a brain magnet in the medical sense — that's TMS, real magnetic pulses applied to the scalp (rTMS is FDA-cleared for depression). Different thing entirely; this is a memory model, not a device.

// sources

Ising 1925 · Little 1974 · Hopfield 1982 (PNAS) · Amit, Gutfreund & Sompolinsky 1985 · Nobel Prize in Physics 2024 (Hopfield & Hinton).