← Back to Home
ON THIS DAY 2026-10-31

1815: Karl Weierstrass is born

On 31 October 1815, in the Westphalian village of Ostenfelde, a boy was born who would spend his early career teaching at a secondary school in rural Prussia before quietly becoming one of the most influential mathematicians of the nineteenth century. Karl Weierstrass is often called the "father of modern analysis," and his life's work was less about discovering new mathematical objects than about making the old ones trustworthy.

Before Weierstrass, calculus ran largely on intuition: limits, continuity, and infinitesimals were explained through pictures and appeals to common sense, which occasionally produced false "proofs." Weierstrass insisted on precision instead, formalizing limits and continuity with the now-famous epsilon-delta definitions. He also constructed a continuous function that is nowhere differentiable — a mathematical object so contrary to geometric intuition that it forced mathematicians to admit pictures alone could no longer be trusted as proof.

This drive for rigor was not eccentric perfectionism; it belonged to a much larger nineteenth-century project, joined by Weierstrass's student Georg Cantor and others, to rebuild mathematics on explicit, checkable logical foundations rather than intuition. That project fed directly into David Hilbert's formalist program, which asked whether all mathematical truths could be derived mechanically from axioms by fixed rules. Hilbert's question provoked Kurt Gödel's incompleteness theorems and, in 1936, Alan Turing's paper defining the abstract machine that bears his name — the theoretical ancestor of every digital computer.

Weierstrass also left artificial intelligence a more direct gift. His approximation theorem showed that any continuous function on a closed interval can be approximated as closely as desired by a polynomial. A century later, this result became the mathematical backbone of the "universal approximation theorem," which explains why artificial neural networks — stacks of simple weighted functions — can, in principle, approximate almost any function at all. It is the theoretical reason deep learning is allowed to work.

So when a modern neural network trains on gradients and convergence criteria, it is quietly running on the same insistence on rigor that the young schoolteacher from Ostenfelde brought to mathematics two centuries ago — proof that today's AI stands on foundations laid long before anyone imagined a thinking machine.