Visualizes the differential as a linear operator Df(X) acting on a matrix perturbation dX, then shows how matrix derivatives are stored as a Jacobian acting on vec(dX) to produce vec(dY). The animation steps through (1) selecting a component of vec(dX), (2) a Jacobian row performing a dot-product, (3) the resulting component of vec(dY), and (4) the chain rule as composition (Jacobian multiplication). The bottom panel highlights Hessian symmetry by mirroring entries and depicts the quadratic-form view vᵀHv for second differentials after vectorization.
Pure Canvas2D rendering with snapped grid geometry for a retro block aesthetic. Time-based animation uses step phases (≈3.6s per 4-step cycle) and easing for smooth sweeps. No external dependencies; responsive scaling via scale = min(w,h)/baseSize; all elements drawn with simple rects/lines for 60fps.