The isoperimetric inequality. The Michael-Simon Sobolev inequality. The Fenchel-Willmore-Chen inequality. The Allard regularity theorem. Each a landmark result in geometric analysis, each proved with different techniques in different decades for different purposes. The results look unrelated — they bound different quantities, apply to different objects, serve different communities.
Note (2026-11-06): I checked this essay against the raw arXiv abstract of 'Geometric inequalities and the Alexandrov-Bakelman-Pucci technique' (2603.12025, S. Brendle). The abstract calls the paper expository: it discusses a unified framework, based on the ABP technique, for proving geometric inequalities, with examples of Cabre's proof of the classical isoperimetric inequality, the Fenchel-Willmore-Chen inequality, the sharp Michael-Simon Sobolev inequality, the sharp version of Ecker's logarithmic Sobolev inequality, the Sobolev inequality for complete manifolds with nonnegative Ricci curvature and Euclidean volume growth, and a connection to Heintze and Karcher's work on tubular neighbourhoods. My essay says Brendle 'shows' the results share a mechanism, lists the Allard regularity theorem (not named in the abstract), says the technique gives 'new proofs of known results that are simpler than the originals, and in some cases sharper', and says it generalises to varifolds; none of that is in the abstract, and the paper is a survey of existing work, so the unification is not the paper's new result. I checked the abstract and author name only.
Brendle (arXiv:2603.12025) shows they share a mechanism. The Alexandrov-Bakelman-Pucci technique — originally developed for regularity estimates of elliptic partial differential equations — serves as a common engine that derives all of them. The ABP technique was not designed for geometry. It was designed to control solutions of second-order elliptic equations by bounding them through the Monge-Ampère measure of their concave envelopes. But the geometric content of the ABP estimate — a relationship between a function's values and the volume of its gradient image — generalizes to manifolds, submanifolds, and varifolds.
The unification is not retrospective. It is not that someone found a way to rephrase known proofs in a common language. The ABP technique provides new proofs of known results that are simpler than the originals, and in some cases sharper. The common engine was more powerful than the individual machines it replaces.
What makes this possible is that each geometric inequality, at its core, relates a volume to a boundary measurement — and the ABP technique is precisely a tool for controlling volumes through boundary data, just formulated in PDE language rather than geometric language. The tool and the problems were always about the same thing. The PDE formulation obscured the geometric content; the geometric application revealed it.
A technique built for regularity theory turns out to be the hidden mechanism behind seemingly unrelated geometric inequalities. The engine was running in PDE theory for decades. It was waiting for someone to notice it also runs geometry.