The junction & the SFQ pulse. Josephson 1962 (Nobel); RSFQ logic (Likharev); SQUIDs in every MRI. The threshold and the single-flux-quantum spike are bedrock — the fastest, lowest-energy switch known.
Superconducting neuromorphic. JJ integrate-and-fire neurons and SFQ neural nets are demonstrated — genuinely fast and efficient, but they live at 4 K. The cryostat is the cost.
The quantum neuron. A junction + capacitor is a transmon qubit; the phase becomes quantum. Quantum perceptrons / large superconducting nets are open — the door this body opens.
A biased junction behaves like a marble on a tilted washboard — the potential U(φ) = −E_J cos φ − (ℏ/2e)·I·φ. The two Josephson relations run the show: I = Ic·sin φ · dφ/dt = 2eV/ℏ Below the critical current the marble is trapped in a well — the phase is pinned, there's zero voltage, a perfect dissipationless supercurrent. Tilt past Ic and the wells vanish: the marble runs downhill, and every well it rolls over slips one flux quantum through the junction — a voltage spike. That escape is the neuron's threshold.
Wire the weighted inputs in as bias current and the junction is an integrate-and-fire neuron with no extra parts: sub-threshold input leaks away as supercurrent; cross Ic and it fires a quantized SFQ pulse, the phase slips by 2π, and it resets — automatically. Every spike is identical and carries exactly one flux quantum. And it does this in picoseconds, dissipating attojoules — the fastest, leanest neuron physics offers.
Quantized, self-resetting, identical spikes — a junction is almost suspiciously well-suited to be a neuron. The catch is the same as all of superconducting computing: it only works cold, and getting signals in and out of the cryostat is its own art.
Same rule, same result — AND/OR in a few sweeps, XOR stuck at 3/4 (verified), freed by a second junction layer. But this body has two endings the others don't. First: add a capacitor and the washboard becomes a quantum well — the phase is now an operator, the lowest two levels are a transmon qubit, and the classical neuron becomes a quantum one. Second: a junction is just two superconductors joined at a weak link — which is literally the ROOT0 <• — so you can build a working one out of folders.
<•Two electrode folders joined at the-dot — the witnessed •. The phase difference φ lives at the dot, because the junction is the dot. Run python evolve.py and it integrates I=Ic·sinφ; push the bias past Ic and sfq-pulses/ fills with files — the neuron firing, counted on disk. browse / clone it →
The folder model isn't a toy analogy — it runs the same equations the silicon does, and it makes <• literal: a structure you can open, with firing you can list. It's the smallest honest junction, and the doorstep of the transmon.