The Takagi function is a classical curiosity: continuous everywhere, differentiable nowhere. Built by summing sawtooth functions at doubling frequencies with halving amplitudes, it produces a fractal curve that wobbles at every scale.
The paper on the dynamics of the Takagi function (arXiv: 2603.22221) treats this function as a dynamical system — iterating it as a map from [0,1] to itself — and discovers that almost every orbit converges to the fixed point 2/3.
This is unexpected. The function's fractal complexity suggests chaotic dynamics. But the chaos lives in the geometry (the graph is nowhere differentiable), not in the iteration (orbits are attracted). The shadowing property — the ability to approximate any pseudo-orbit with a true orbit — holds for the original Takagi function. But for scaled versions T_γ = γT, there exist parameter values where shadowing breaks. The function's geometric complexity and its dynamical simplicity coexist without contradiction.
The through-claim: geometric complexity and dynamical simplicity are independent axes. A function can be maximally irregular in its shape (nowhere differentiable) while being maximally regular in its iteration (globally attracting fixed point). The fractal structure is a property of the graph; the convergence is a property of the map. Complexity in one sense does not predict complexity in the other.
2603.22221. Dynamical systems / Takagi function / shadowing property / nowhere differentiable functions / fixed point attraction.