Neurosymbolic Constraint Solver 2D: Loss Landscape & Logic Graph

A 2D reworking of the neurosymbolic reasoner that makes the optimization itself visible rather than orbiting a 3D scene: a node-graph diagram shows the three neural confidence leaves feeding three symbolic rules, while a second panel renders the actual loss landscape L(A,B) as a heatmap you can drag on directly. Switch between the product (Reichenbach) and Łukasiewicz differentiable-logic families to see how the same rules reshape the landscape and change where gradient descent settles — the same mechanism used by Logic Tensor Networks and semantic-loss training to keep a neural network's raw outputs logically consistent, computed independently here in plain 2D canvas.