P(|1⟩)=sin²(θ/2). Every other body had to find its threshold. The cubi's threshold is the wavefunction collapsing.The cubi neuron itself. A single qubit rotated and measured — angle encoding, the Born-rule activation — is exact and runs on real quantum hardware (Tacchino et al. ran a quantum perceptron on IBM Q, 2019).
Variational quantum classifiers. Small quantum neural nets are demonstrated, but it's the NISQ era — noise, decoherence, and barren plateaus (vanishing gradients) make training hard at scale.
Quantum advantage for learning. Whether a quantum model actually beats a classical one on real data is open and genuinely contested — the honest line between physics and hype.
A cubi's state is a point on the Bloch sphere: north pole is |0⟩, south is |1⟩, and everywhere between is a superposition α|0⟩ + e^{iφ}β|1⟩. The weights and inputs are rotations — they turn the arrow. But you never read α and β directly; you measure, and the state collapses to |0⟩ or |1⟩ with probability P(|1⟩) = sin²(θ/2) That curve — 0 at the north pole, ½ at the equator, 1 at the south — is a built-in nonlinear activation. The measurement is the neuron.
Encode the inputs as rotation angles and the weights as how hard each one turns the cubi. The rotations compose — the state ends at θ = b + Σ wᵢxᵢ — and the Born rule reads out the activation P(|1⟩) = sin²(θ/2). Threshold it (more likely |1⟩ than |0⟩) and the cubi fires. It's the same weighted-sum-then-squash as every body — except the squash is physics you can't avoid.
Train the rotation angles and the cubi learns AND/OR in a few sweeps. One cubi, encoded linearly, is still one cut — XOR stalls at 3/4 (verified), exactly like its seven classical cousins. But the quantum world has a different escape than stacking layers: a feature map that entangles the inputs adds an x₁·x₂ cross-term, lifting the data into a space where XOR is separable. Flip ⊗ entangle and XOR snaps to 4/4. Entanglement does what depth did — that's the quantum twist on the oldest wall in the series.
The honest line (this is the contested one). The cubi's mechanism is real and runs on real hardware. The feature map lifting XOR is exactly how quantum kernels work, and it's genuinely quantum. But whether this buys a real-world advantage over a classical net — once you account for noise, the cost of loading data into amplitudes, and reading it back out — is unsettled and hotly argued. The physics is sound; the supremacy is a promissory note. We mark it FRONTIER on purpose.
This is where the road that began in a branch predictor arrives: the dot product turned into a quantum amplitude, the threshold into a measurement, depth into entanglement. Same neuron, eighth body — and the first one that is genuinely uncertain about its own answer until you look.