The Ground Beneath Math

This is the unsettling thing Alexander Grothendieck spent roughly two decades demonstrating, systematically and without apparent effort, before walking away from the whole edifice he'd built. His departure is almost as consequential as his work — if only because it forces the question of what you do once you've seen that far.

Mathematics before Grothendieck operated largely at the level of objects. Here is a curve. Here is a surface. Here is an equation; find its solutions. The implicit assumption was that mathematical objects were the things worth studying, and that progress meant discovering more facts about them. Grothendieck worked like someone who'd realized the objects were symptoms. He wanted the underlying structure. Then the structure beneath the structure. Then whatever that was resting on.

He arrived at the Institut des Hautes Études Scientifiques outside Paris in 1958, born stateless — his mother German, his father a Russian-Ukrainian anarchist who was murdered at Auschwitz. Grothendieck had spent parts of his childhood hiding in southern France, classified as an "undesirable alien." He had no permanent nationality for much of his adult life. He was always outside the system he was working within.

By the time he was thirty, he had already solved major problems in functional analysis and pivoted to algebraic geometry — a discipline concerned with the relationship between polynomial equations and geometric shape. Within a decade, he had not improved it or extended it. He had rebuilt it from scratch, with different materials, on a different foundation, using a completely different conception of what "space" even means.

The result was a mathematics that could see things its predecessors couldn't, applied to problems its predecessors couldn't properly formulate. And the method he used offers, for anyone paying attention, a precise demonstration of what it means to understand something rather than merely compute it.

i · the rising sea

Grothendieck described his approach once with an image that has since become canonical among mathematicians: rather than attacking a hard problem directly — hammering at a nut with a chisel — you let the sea rise until the nut floats. This is not a metaphor for patience. It is a description of the correct level of abstraction.

Classical algebraic geometry studied varieties: sets of points in space where polynomial equations hold. The variety defined by x² + y² = 1 is a circle. Clean, geometric, and it works — until you try to study equations over finite fields, or want to count integer solutions, or need to connect arithmetic to geometric intuition, at which point the framework starts leaking. The objects don't generalize cleanly. The tools don't transfer. You end up patching.

Grothendieck's solution was to change what you're looking at. Instead of studying the set of points where an equation holds, study the ring of functions on that set. And instead of working with one ring at a time, work with the entire category of rings and all the maps between them. This is the concept he called a scheme: a geometric object built not from points but from algebra, from the structure of how different algebraic systems relate to each other.

The shift — from objects to relationships between objects — turned out to be extraordinarily generative. When you place the right framework around a problem, the hard problem doesn't get solved by cleverness. It dissolves. What was previously a difficult technical challenge becomes an obvious consequence of the structure. The sea rose; the nut floated; the problem was never as hard as it looked from the wrong vantage point.

A striking example: the Weil conjectures, formulated in 1949, were a set of predictions about counting solutions to equations over finite fields. They were considered extremely difficult. They required an analogue of topology — a way to count "holes" in geometric objects — that simply didn't exist over finite fields. Grothendieck invented it: étale cohomology, a completely new cohomological theory built from first principles in the new language of schemes. The Weil conjectures, once impossible to approach, became approachable. His student Pierre Deligne proved the last of them in 1974.

This is the rising sea method in action. You don't solve the problem directly. You build a sea large enough that the problem floats.

The approach didn't stay in algebraic geometry. Category theory — the branch of mathematics that formalizes the study of relationships and structure rather than objects themselves — now appears throughout theoretical computer science, logic, quantum physics, and linguistics. The functors and natural transformations Grothendieck developed to talk about geometric spaces turned out to be a vocabulary for talking about structure itself, portable across every domain where structure exists. He solved an algebraic geometry problem and accidentally provided a language for half of mathematics.

ii · the shape of logic

The pattern is recognizable beyond mathematics: clarity about what kind of problem you're solving dissolves effort that would otherwise be wasted on the wrong tools. Friction is a signal. You're pushing on the wrong surface.

The deepest discovery in Grothendieck's trajectory was the topos.

A topos is a category that behaves like a geometric space in a precise technical sense — it has a logic built into it. The extraordinary thing, the thing that produces genuine vertigo, is that this logic is not necessarily classical. In a topos, you can have geometries where the law of excluded middle doesn't hold. Spaces where a proposition can be "not true" without being false. Mathematical universes with their own internal rules about what can be known and proven.

Grothendieck's work forces an implication: geometry and logic are not separate branches of mathematics with occasional intersections. They are the same thing viewed from different angles. A geometric space is a logic. A logic is a space. The "shape" of a set of axioms is literal: it is a geometry. When you choose a foundation for mathematics — when you choose which axioms to work with — you are choosing a geometry. When you study a space, you are studying a logic.

This result, developed by Grothendieck and extended by William Lawvere, Myles Tierney, and others, means that the relationship between what exists and what can be said about it has a structure, and that structure determines what counts as true. Mathematical truth is not absolute in the way we pretend it is. It's relative to the topos. This is not philosophical wooliness — it's a theorem.

Classical mathematics takes place in one particular topos: the category of sets, with classical logic, where every proposition is either true or false. Grothendieck's framework revealed this as one option among many. Other toposes have different truths, different valid inferences, different notions of what "exists" means. The foundations aren't bedrock. They're a choice.

Here's the vertigo: Grothendieck didn't approach the foundations of mathematics and find solid ground. He found that what looked like ground was floating. And then he mapped part of the ocean beneath it.

This is not a mystical claim. It is precise mathematics. The precision is what makes it vertiginous. The clearer the picture becomes, the less stable the footing feels — not because the mathematics is wrong, but because it is exactly right.

iii · walking away

In 1970, Grothendieck learned that the IHES received partial funding from the French military. He resigned immediately and permanently.

This was not a symbolic protest. The IHES was arguably the best place on Earth to do the kind of mathematics he was doing. He had the infrastructure, the collaborators, the momentum of a decades-long project — the Séminaire de Géométrie Algébrique, which had by then produced thousands of pages of new foundations. He abandoned it anyway. He spent the following years working on ecology and pacifist activism, editing a journal he called Survivre (Survive), writing about nuclear proliferation and environmental collapse.

He returned to mathematics once more, briefly, then left again, more completely. In the 1980s, he wrote a massive manuscript called Récoltes et Semailles — Reapings and Sowings — part autobiography, part mathematical meditation, part sustained accusation. He accused his former students and colleagues of intellectual theft, of misrepresenting his legacy, of doing mathematics for prestige rather than understanding. The manuscript circulated in partial form, in French, for decades, never officially published. A full edition appeared only recently.

By 1991, he had moved to a remote village in the Pyrenees and withdrawn from all contact. He wrote tens of thousands of additional pages — philosophical, spiritual, cosmological — that remain largely unread. He died in November 2014. The location was not widely known until after his death.

The question his withdrawal poses is serious. What do you do once you've rebuilt the foundations? Grothendieck's departure looks, from one angle, like burnout or bitterness. From another, it looks like someone who had seen something clearly and found it impossible to pretend the professional apparatus was the important thing. When you've demonstrated that the shape of logic is literally a geometry, that the ground beneath math is not itself mathematical, it becomes difficult to care deeply about conference proceedings and citation counts.

The renewed attention his work receives now — the retrospectives, the ongoing translation projects, the spreading influence of category theory from physics to programming language design — suggests that mathematics is still catching up to what he built five decades ago. The conceptual infrastructure he laid is still being discovered, still being applied to problems he probably would have said it was designed for. Homological algebra in data science. Topos theory in quantum gravity. Functors in type systems.

The sea he let rise is still rising. The problems it was built to dissolve are still being recognized as problems it can dissolve.

Whether that's reassuring depends on your tolerance for depth. The void at the center of Grothendieck's legacy is not his disappearance into the Pyrenees. It's the opening he left: the discovery that the ground beneath structure is more structure, all the way down, and that the structures determine what's true, and that there's no final level, and that this is fine. More than fine. It's the actual situation.

He mapped part of it, then walked into the mountains.

The mathematics remains. The nut is still floating.

iv · sources

• How Alexander Grothendieck Revolutionized 20th-Century Mathematics — Quanta Magazine, 2026-05-20
• A Mad Day's Work: From Grothendieck to Connes and Kontsevich — Notices of the AMS, 2001-09
• The Rising Sea: Grothendieck on Simplicity and Generality — Colin McLarty, 2003