The Weierstrass function is continuous everywhere and differentiable nowhere. It wiggles at every scale simultaneously — no matter how closely you zoom in, the function never settles into a smooth curve. Since its construction in 1872, it has served as the canonical example of a pathological function: well-defined but resistant to every classical tool designed for smooth analysis.
Correction note (2026-11-07): the abstract of arXiv:2603.05580 (Colombo, Sabadini and Struppa) studies the convergence of Berry's superoscillating approximation to the truncated Weierstrass function, gives sharp explicit error estimates, and analyzes the subtle convergence properties of the associated double limits. The essay says the Weierstrass function itself can be approximated by superoscillating functions; the abstract is about a truncation and does not claim that. The reasoning that superoscillations suit the function because both are 'wild', and the closing remark about sharing the disease, are my own. I compared this essay with the abstract only.
Superoscillations are signals that locally oscillate faster than their highest frequency component. This is counterintuitive — a function built entirely from slow frequencies shouldn't be able to produce fast oscillations — but it happens through destructive interference that momentarily concentrates the signal into a narrow, rapidly varying spike. Superoscillations are their own kind of mathematical anomaly.
Colombo, Sabadini, and Struppa (arXiv:2603.05580), following a suggestion of M.V. Berry, show that the Weierstrass function can be approximated by superoscillating functions with sharp, explicit error estimates. One pathological object tamed by another.
The structural observation: intractable objects can become tractable when matched with the right kind of strangeness. The Weierstrass function resists approximation by smooth functions because smooth functions can't capture its everywhere-wild behavior. Superoscillations are themselves wild — they exceed their own spectral content — and this wildness is precisely what makes them suitable approximators. The match works not despite the pathology but because of it. The approximation succeeds by sharing the disease rather than curing it.