Mathematics after Grothendieck

Not a kingdom with one heir

There is no moment at which mathematics passes from a dead mathematician to a designated successor. A theorem keeps its proof. A definition can migrate into textbooks until speakers use it without recalling the paper in which it was named. A program may divide into problems whose solutions require methods the proposer never knew. A metaphor may survive while the technical object beneath it is replaced. “Grothendieck’s legacy” is therefore not a territory owned by everyone who works inside it, nor a genealogical claim that all modern algebraic geometry descends from one mind.

The concrete task is smaller and more revealing: trace paths from a documented construction or program associated with Grothendieck to mathematics practiced after his death. A path needs intermediate names, texts, and changes of meaning. It should show where a later mathematician solved a new problem rather than credit the result backward. It should also allow a negative answer. Similar notation does not prove influence; a broad concern with structure is not uniquely Grothendieckian; and abstraction itself has many histories.

Six paths stand out because each can be followed through working references rather than commemorative adjectives. Schemes became ordinary infrastructure. Sites, sheaves, descent, and the six operations traveled into new geometries. Stacks acquired higher and derived forms. Motives developed into several rigorous theories while retaining conjectural horizons. The surface pictures of Esquisse d’un Programme opened research around dessins, Galois actions, Teichmüller structures, and anabelian reconstruction. Topoi moved not only through geometry but through higher category theory and categorical logic. These routes cross, and none is finished. 11The Stacks Project Authors, The Stacks Project, 1; The Stacks Project Authors, The Stacks Project; Olivia Caramello, Theories, Sites, Toposes: Relating and Studying Mathematical Theories Through Topos-Theoretic “Bridges” (Oxford University Press, 2018), chs. 1–4.

The archive matters to this history, but it is not its sole engine. Mathematicians were reading seminar volumes, letters, circulated typescripts, published surveys, and one another’s papers long before Montpellier put scans online. Conversely, the fact that a late note resembles a modern theory does not prove that the note caused it. Influence can be direct, mediated, parallel, or retrospectively recognized. This chapter uses the modest verbs provided a language, posed a program, was developed by, and is used in. It avoids the imperial verb contains, by which one great precursor is made to contain the future.

First path: schemes become infrastructure

The human problem behind a scheme is continuity across kinds of equation. Classical algebraic geometry studied polynomial solutions over fields, often with geometric intuition drawn from complex numbers. Arithmetic asks about the same formulas over integers, finite fields, local fields, and families of rings. If every setting requires a new foundational object, comparison becomes fragile. Scheme theory provides a common arena in which arithmetic and geometric fibers belong to one relative construction.

What a scheme says

For a commutative ring $A$, the affine scheme $\operatorname{Spec} A$ consists of its prime ideals, equipped with the Zariski topology and a structure sheaf that records local algebra. A scheme is a locally ringed space covered by open pieces isomorphic to affine schemes. Given maps $X\to S$ and $S'\to S$, the fiber product $X\times_S S'$ expresses change of base. This definition is compact; its consequences include nilpotent structure, gluing, arithmetic fibers, moduli constructions, and cohomology. 22The Stacks Project Authors, The Stacks Project, 1; The Stacks Project Authors, The Stacks Project.

The scheme $\operatorname{Spec}\mathbb Z$ is a concrete anchor: its closed points correspond to prime numbers, while its generic point corresponds to the zero ideal. Base change to $\operatorname{Spec}\mathbb F_p$ isolates the fiber at $p$. This example has no faithful picture as a curve in ordinary space; that failure marks the boundary of the geometric analogy.

The first intuition is a family. A scheme over a base $S$ can organize objects that vary as the base varies. The limit of the analogy is that neither base nor fiber needs to be an ordinary space, and nilpotent functions can carry information with no visible point-set shadow. Saying “a scheme is a generalized shape” is harmless only until it encourages the reader to expect a picture for every prime ideal and every infinitesimal thickening.

After Grothendieck, the strongest sign of survival is not repeated praise but routine use. The Stacks Project builds commutative algebra, schemes, algebraic spaces, and algebraic stacks into an open, continuously revised reference. It identifies results by permanent tags, records contributors, and accepts corrections through a maintainer. The project began as collaborative notes, placed a first website online in 2012, and continues to grow. This is not EGA transferred to a browser. It is a different collective object with different authorship, scope, notation, and technical choices. Yet it shows how the scheme language became infrastructure: researchers need a durable route through lemmas whose dependencies can be followed and repaired. 33The Stacks Project Authors, The Stacks Project.

That digital form also changes authority. EGA’s architecture and the SGA seminars emerged from a concentrated school, with Grothendieck, Dieudonné, seminar speakers, editors, and students occupying unequal but visible roles. The Stacks Project distributes correction and extension across a wider contributor network while retaining a maintainer. Neither model is authorless. Both demonstrate that a foundational language survives through organized labor, not through the charisma of its first builder.

Second path: locality is chosen

An ordinary topological space has specified open sets. Many problems in algebraic geometry demand a different notion of local observation. Polynomial equations may become simpler after an étale map rather than after restriction to a Zariski-open subset. Descent asks whether objects constructed on such local tests, together with compatible identifications, come from an object over the original base. Grothendieck’s language of sites separates locality from the topology of points.

Site, sheaf, topos

A site is a category equipped with families of morphisms declared to be coverings, subject to identity, refinement, and base-change conditions. A presheaf assigns data contravariantly to objects of the category. It is a sheaf when compatible local sections glue uniquely. A Grothendieck topos is a category equivalent to the category of sheaves of sets on some site. Different sites can present equivalent topoi, so the presentation and the represented “world of variable sets” are related but not identical. 44The Stacks Project Authors, The Stacks Project; Jacob Lurie, Higher Topos Theory, Annals of Mathematics Studies 170 (Princeton University Press, 2009), 170; Caramello, Theories, Sites, Toposes: Relating and Studying Mathematical Theories Through Topos-Theoretic “Bridges”, chs. 1–4.

The open subsets of an ordinary topological space, with ordinary open covers, form a site; ordinary sheaves on the space are sheaves on that site. The example is genuine but limited. Étale covers of schemes need not be inclusions of open subsets, which is precisely why the abstract definition matters.

The intuition is an agreement about tests. To understand an object, the theory specifies which probes count as local and which families count as sufficient coverage. The limit is severe: a topos cannot be reduced to an ordinary space with exotic open sets. It is a category with logical and geometric structure, and equivalence of topoi can erase visible differences between site presentations.

This idea survived through cohomological practice. Étale cohomology, descent, and the formalism of operations became standard equipment across arithmetic geometry. The “six operations” name a web of pullback, pushforward, exceptional pullback and pushforward, tensor product, and internal Hom, joined by adjunctions and compatibilities. No single slogan captures the hypotheses required in each setting. What persists is the expectation that a good cohomology theory should travel functorially with spaces and maps.

The modern theory of diamonds provides a concrete, post-2014 view of that persistence. Peter Scholze’s Étale cohomology of diamonds, first posted in 2017 and revised through a fourth arXiv version in April 2026, develops a six-functor formalism for diamonds and small $v$-stacks on perfectoid spaces of characteristic $p$. Its abstract says that a natural functor from analytic adic spaces over $\mathbb Z_p$ to diamonds identifies étale sites and recovers an earlier formalism in the noetherian setting. The theory and its theorems are Scholze’s; the older vocabulary of sites, descent, and six operations establishes a structural lineage, not transferred authorship. 55Peter Scholze, “Étale Cohomology of Diamonds,” Preprint, arXiv:1709.07343, 2017, https://arxiv.org/abs/1709.07343; The Stacks Project Authors, The Stacks Project.

The image of a diamond is especially dangerous for a general reader. A diamond is not a faceted solid. Roughly, it is a sheaf for the very fine $v$-topology that can be presented as a quotient of a perfectoid space by a suitable equivalence relation. The point of changing the topology is not decorative generality. It allows geometric and cohomological objects to exist in a setting where ordinary schemes or adic spaces do not supply enough flexibility. The old decision to axiomatize coverings makes the new decision possible; the new mathematics lies in determining which coverings, quotients, finiteness conditions, and cohomological theorems actually work.

Every perfectoid space represents a $v$-sheaf and hence gives a basic diamond. Not every $v$-sheaf is a diamond: the quotient presentation and size conditions carry mathematical content. The example anchors the definition without reducing diamonds to renamed perfectoid spaces. 66Scholze, “Étale Cohomology of Diamonds”; The Stacks Project Authors, The Stacks Project.

Third path: symmetries refuse to disappear

Moduli problems classify objects: curves, bundles, maps, representations, and much else. A set of isomorphism classes throws away automorphisms. Two objects can occupy the same coarse point while carrying different symmetry, and families can fail to glue if the identifications on overlaps are treated as mere equality. A stack retains those automorphisms and makes descent part of the object’s definition.

Why a stack is not a pile

A stack in groupoids over a site assigns a groupoid to each test object. Isomorphisms between two local objects form a sheaf. Compatible local objects and coherent isomorphisms on overlaps must descend effectively to a global object. The relevant Stacks Project definition is Tag 02ZI. “Groupoid” means that all arrows are invertible; “effective” means the local descent datum actually comes from a global object instead of only satisfying written compatibility equations. 77The Stacks Project Authors, The Stacks Project, 1; The Stacks Project Authors, The Stacks Project.

The moduli stack of elliptic curves supplies a true example. An elliptic curve carries automorphisms, including the inversion map in the usual setting; the coarse $j$-line records an isomorphism class but forgets that stabilizer. The stack retains it. This is structural information, not an extra label attached after classification.

The faithful intuition is a moduli space that remembers stabilizers and gluing. The misleading intuition is a set whose points have extra labels. A stack is categorical through and through. Its symmetries affect geometry, cohomology, and deformation. This is why stack language became routine in moduli theory rather than remaining a philosophical ornament.

Grothendieck’s 1983 Pursuing Stacks pressed beyond ordinary categorical levels. Brown’s account traces its prehistory to their 1982 mathematical correspondence and to three letters Grothendieck sent Lawrence Breen in 1975. Brown describes the letter to Daniel Quillen as a prelude to Pursuing Stacks proper, not as the manuscript’s sole origin. The sequence matters because correspondence generated the manuscript, whereas Quillen’s name identifies only one textual threshold within its later architecture. The work developed a program around homotopical algebra, models of homotopy types, and higher-categorical coherence. A revised electronic edition appeared on arXiv in 2021. Its path into modern higher categories passed through many mathematicians and through models and theorems not supplied by the 1983 text. 88Grothendieck, À la Poursuite des Champs (Pursuing Stacks); Grothendieck, Pursuing Stacks (À La Poursuite Des Champs), vol. 1, vol. 1.

Derived algebraic geometry is one major destination. Ordinary intersection can lose information when two subspaces fail to meet transversely. Algebraically, higher Tor groups detect the failure. A derived fiber product retains that homotopical data instead of truncating it away. More broadly, derived schemes and stacks organize deformation problems, cotangent complexes, and moduli with coherent higher structure. Bertrand Toën’s survey describes the field and its connections to deformation quantization; Jacob Lurie’s higher topos and higher algebra frameworks provide another extensive foundation. 99Bertrand Toën, “Derived Algebraic Geometry,” EMS Surveys in Mathematical Sciences 1, no. 2 (2014): 153–240, https://arxiv.org/abs/1401.1044; Lurie, Higher Topos Theory, 170; Grothendieck, À la Poursuite des Champs (Pursuing Stacks).

The self-intersection of a line in the affine plane is a concrete anchor. The ordinary fiber product returns the line and loses the failure of transversality; the derived fiber product retains a nonzero higher Tor contribution. Derived structure records that excess intersection, but it does not make every intersection calculation easy.

Describing Grothendieck as having predicted derived geometry is too smooth. He helped formulate derived-category methods, promoted stacks, and in Pursuing Stacks envisioned a synthesis of homotopy and higher categorical structure. Those are substantial, documentable antecedents. Modern derived algebraic geometry nevertheless depends on later solutions to coherence, model-independence, representability, and computation. A program can be forward-looking without being proprietary. Giving Toën, Vezzosi, Lurie, and other builders their verbs—defined, proved, constructed—does not reduce Grothendieck’s vision. It makes the path accurate.

Fourth path: the plural of motive

Grothendieck’s idea of motives began from a striking pattern. Algebraic varieties admit several cohomology theories: Betti, de Rham, $\ell$-adic, crystalline, and others in suitable settings. These theories produce related invariants and comparison maps. The program sought a universal kind of object from which particular cohomologies could arise as realizations. The word motive, chosen from the idea of an underlying motif, names that aspiration.

What is and is not constructed

There are rigorous categories of pure motives under chosen equivalence relations and coefficient systems, triangulated categories of motives, motivic cohomology, and stable motivic homotopy categories. There also remain conjectural structures, including expected categories and $t$-structures with properties not established in full generality. It is therefore inaccurate to speak as if “the category of motives” were one finished edifice. 1010Thomas Geisser et al., “Motives and Homotopy Theory of Schemes,” Oberwolfach Reports 1 (2004), https://ems.press/journals/owr/articles/698; International Mathematical Union, “2002 Fields Medals and Nevanlinna Prize: Vladimir Voevodsky,” 2002, https://www.mathunion.org/fileadmin/IMU/ICM2002/offline/Beijing/Spot/Fields_2002.htm. Thomas Geisser, “Motives and Homotopy”, Oberwolfach Reports 1, no. 2; IMU, “Voevodsky Fields Citation”, official 2002 citation.

In a standard category of pure motives, the projective line decomposes into a unit motive and a Lefschetz motive. Its familiar cohomology in degrees zero and two reflects those two pieces under realization. The example is precise only after coefficients and the category of motives are specified; it cannot certify a universal mixed-motive category.

The intuition is musical: several cohomology theories sound different realizations of shared underlying material. Its limit is that a metaphor does not construct the universal object, and different rigorous theories answer different universal problems. Even the term realization carries technical functors and compatibility conditions that cannot be supplied by the image of a tune.

Vladimir Voevodsky, in a field also shaped by Fabien Morel and many others, developed motivic cohomology and $\mathbb A^1$-homotopy methods used in major proofs. The International Mathematical Union’s 2002 Fields Medal citation recognizes his development of new cohomology theories for algebraic varieties and the resulting insight into number theory and geometry. The lineage therefore has separate clauses: Grothendieck articulated a universal cohomological program; Voevodsky and later collaborators supplied distinct constructions and proofs that the earlier program did not contain as completed arguments. 1111International Mathematical Union, “2002 Fields Medals and Nevanlinna Prize: Vladimir Voevodsky”; Geisser et al., “Motives and Homotopy Theory of Schemes”. IMU, “Voevodsky Fields Citation”, official 2002 citation; Thomas Geisser, “Motives and Homotopy”, Oberwolfach Reports 1, no. 2.

Program, construction, and proof remain distinct: Grothendieck receives credit for documented questions and languages, while later mathematicians retain their own documented results and separate authorship.

Motivic homotopy offers another example of changed meaning. The affine line $\mathbb A^1$ plays an interval-like role in defining homotopy for schemes. But an algebraic interval is not the unit interval of ordinary topology, and invariance under $X\times\mathbb A^1$ introduces arithmetic and algebro-geometric phenomena with no simple visual equivalent. The achievement lies in the precise category, topology, stabilization, and representability theorems, not in substituting one interval symbol for another.

The posthumous legacy here is a productive incompletion. Motives name successes and open conjectures at once. They connect algebraic cycles, $K$-theory, cohomology, homotopy, and arithmetic. No newly opened box in Montpellier needs to contain a final theory for the program to have mattered. Its historical force is already visible in the communities, constructions, and unresolved questions that took it forward.

Fifth path: a child’s drawing and the absolute Galois group

In Esquisse d’un Programme, Grothendieck described surface topology as leading him back to algebraic geometry through objects simple enough for a child to encounter in play. The phrase became attached to dessins d’enfants: bipartite graphs embedded on oriented compact surfaces, connected through Belyi maps to algebraic curves defined over number fields. 1212Grothendieck, “Esquisse d’un Programme,” 7–48; Grothendieck, “Esquisse d’un Programme,” 003.

The dessin correspondence

A Belyi map is a finite morphism from a smooth projective algebraic curve to the projective line, ramified only above $0$, $1$, and $\infty$. The inverse image of the interval from $0$ to $1$ gives an embedded bipartite graph: preimages of $0$ and $1$ supply the two colors of vertices, while local ramification controls valencies. Belyi’s theorem relates the existence of such a map to definition over the algebraic closure of $\mathbb Q$. The absolute Galois group acts on the coefficients and hence on dessins; the action on their isomorphism classes is faithful. 1313Schneps and Lochak, Geometric Galois Actions 1, 242; Grothendieck, “Esquisse d’un Programme,” 7–48; Grothendieck, “Esquisse d’un Programme,” 003; Bibliothèque nationale de France and Société mathématique de France, “Les mille et une pages mathématiques d’Alexandre Grothendieck”.

The map $z\mapsto z^n$ on the projective line is a simple Belyi map: its ramification lies over $0$ and $\infty$, a subset of the allowed three points. The inverse image of $[0,1]$ is a star-shaped dessin. Simplicity of this example does not make general Galois orbits easy to determine.

The human appeal is immediate. A finite drawing carries arithmetic information, so a group famous for its invisibility acts on combinatorial things one can put on paper. The limit is equally important. The graph does not make the absolute Galois group transparent. Determining orbits, stabilizers, and arithmetic fields can be hard, and visually similar drawings can represent different structures once embedding and cyclic order are included.

The Esquisse places dessins within a larger program: moduli spaces of curves, the Teichmüller tower, profinite fundamental groupoids, the action of $\operatorname{Gal}(\overline{\mathbb Q}/\mathbb Q)$, and a hoped-for algebraic characterization of that action. Later Grothendieck-Teichmüller theory, work on profinite braid groups, and extensive study of dessins developed parts of this landscape. The 1997 volume edited by Leila Schneps and Pierre Lochak both published the Esquisse in a scholarly setting and assembled research around it. 1414Schneps and Lochak, Geometric Galois Actions 1, 242.

Anabelian geometry follows a related but distinct question: whether suitable arithmetic varieties can be reconstructed from their profinite fundamental groups together with Galois action. The word anabelian signals objects far from abelian: the noncommutative structure of the fundamental group is meant to retain geometry that abelian invariants lose. Major later theorems have precise hypotheses on curves, fields, and homomorphisms. It would be false to summarize them as “every space is its group.” 1515Grothendieck, “Esquisse d’un Programme,” 7–48; Grothendieck, “Esquisse d’un Programme,” 003; Schneps and Lochak, Geometric Galois Actions 1, 242.

The thrice-punctured projective line, $\mathbb P^1\setminus\{0,1,\infty\}$, is a central anchor because its profinite fundamental group carries a faithful outer Galois action. The reconstruction claims remain dependent on the curve and base field; the example is not a theorem about arbitrary spaces.

The 2025 BnF lecture used dessins as an entrance into the mathematical pages of Esquisse. That public choice shows another form of legacy: an object can mediate between a primary text, specialist research, and explanation to a wider audience without being reduced to a mascot. The child’s drawing is not proof by innocence. It is a combinatorial doorway whose arithmetic room remains demanding. 1616Bibliothèque nationale de France and Société mathématique de France, “Les mille et une pages mathématiques d’Alexandre Grothendieck”.

Sixth path: topoi beyond geometry

Topos theory crossed another boundary into logic. A classifying topos can represent models of a geometric theory across varying topoi. Because one topos can have different site presentations, information expressed naturally in one presentation may be transported into another. Olivia Caramello’s program develops such “topos-theoretic bridges” systematically. This work is neither a footnote to SGA nor a claim that all theories are secretly identical. It is a later research architecture with its own definitions and results. 1717Caramello, Theories, Sites, Toposes: Relating and Studying Mathematical Theories Through Topos-Theoretic “Bridges”, chs. 1–4; The Stacks Project Authors, The Stacks Project.

For a concrete presentation change, sheaves on a topological space can be recovered from a suitable basis of open sets rather than from every open set. The two sites present equivalent topoi. Caramello’s bridges are much more structured, but this elementary case shows why presentation and invariant cannot be identified.

The intuition is that a sufficiently rich invariant can serve as a bridge between shores described in different languages. The limit is that equivalence of classifying topoi does not give a word-for-word dictionary between arbitrary statements. Only suitable invariant properties travel, and proving that a theory has the required presentation is mathematical work.

Higher topos theory moves in another direction. In Lurie’s formulation, infinity-categories replace hom-sets by spaces of morphisms with coherent higher homotopies, and infinity-topoi generalize the sheaf worlds of ordinary topos theory. Again, the path runs through Grothendieck’s sites and topoi and through homotopy-theoretic developments by many authors. It would be inaccurate both to deny the ancestry and to say SGA 4 already contained Higher Topos Theory. 1818Lurie, Higher Topos Theory, 170.

These examples reveal a characteristic of foundational concepts. Their descendants may live far from the problem that first called them forth. A site invented to make étale cohomology possible can become part of $p$-adic geometry, higher geometry, or logic. That mobility is not proof that the concept explains everything. It is evidence that the definition found a stable balance between generality and usable constraints.

How to credit a lineage

The paths above correct two opposite legends. The first imagines that Grothendieck left mathematics in 1970 and therefore has no relation to later stacks, homotopy, or arithmetic programs. The documents disprove the premise: late manuscripts posed sustained mathematical questions. The second legend imagines that he foresaw every important later theory and that later mathematicians merely filled in details. The theorems disprove that conclusion: new foundations, counterexamples, computations, and proofs changed the questions themselves.

A practical credit rule follows. Name Grothendieck where a concept, text, or program can be located. Name collaborators and predecessors where the historical object was collective: Dieudonné in EGA, SGA speakers, Giraud and Artin in stacks and descent, Belyi for his theorem, Quillen and other homotopy theorists around Pursuing Stacks. Name later builders for later mathematics: Voevodsky and Morel, Toën and Vezzosi, Lurie, Scholze, Caramello, and the many authors a short account cannot exhaust. Then state the relation—generalization, solution, reformulation, independent construction—rather than replacing it with a family tree.

The same rule handles the archive. If an unpublished page proves to contain an earlier formulation, cite the cote and date, compare exact statements, and ask whether it circulated. Do not accuse a later author through chronology alone. If a modern paper explicitly cites a manuscript, the path is direct. If practitioners learned the language through other practitioners, the path is communal. If resemblance appears only after the fact, call it a resonance until stronger evidence exists.

Mathematical legacy is most concrete in verbs performed by other people. They teach a definition, correct a proof, extend a formalism, construct an example, disprove a conjecture, build a reference, or return to a program with new tools. In that sense the Stacks Project is as telling as a medal, a $v$-stack as telling as a memorial lecture, and a disputed conjecture as telling as a solved one. The inheritance is not agreement with Grothendieck. It is the capacity to use, test, alter, and sometimes refuse what his work made available.

The result has no center from which all paths can be surveyed. Schemes are infrastructure; motives retain conjectural horizons; dessins are concrete but arithmetically elusive; higher stacks demand formidable coherence; topoi bridge some theories and not others. This unevenness is the truth of a living legacy. It prepares the political issue of who may circulate the pages in which ideas were first or last expressed, and under whose authority.

Four tests for a mathematical lineage

A lineage claim becomes stronger as it passes four tests. The first is textual: the earlier and later sources formulate recognizably related objects or questions. The second is historical: there is evidence that the later author read, heard, or inherited the earlier formulation through an identifiable route. The third is mathematical: the relation survives expert comparison of definitions and hypotheses. The fourth is counterfactual modesty: the claim does not imply that no other route to the later work was possible.

The path from schemes to the Stacks Project passes the textual and mathematical tests immediately. The reference defines schemes in the inherited language and builds the foundations needed for algebraic stacks. Historical transmission is diffuse because generations of books, courses, and papers lie between EGA and an online tag. That diffusion is itself the lineage: scheme language became a prerequisite rather than an exotic citation. 1919The Stacks Project Authors, The Stacks Project; The Stacks Project Authors, The Stacks Project, 1.

The path from Pursuing Stacks to modern higher geometry is different. Textual kinship appears in homotopy types, higher categorical models, and stacks. Historical accounts name the manuscript as programmatic. Mathematical comparison reveals both anticipations and enormous later changes in machinery. A careful sentence can call it an ancestor or source of problems. It should not say that the modern foundations are merely a transcription of Grothendieck’s text. 2020Grothendieck, À la Poursuite des Champs (Pursuing Stacks); Lurie, Higher Topos Theory, 170.

The path from sites and six operations to diamonds is structural rather than a claim of direct personal inspiration. Scholze’s abstract explicitly speaks the language of étale sites, $v$-stacks, and a six-functor formalism. Those terms have technical histories traceable through Grothendieck’s school. The new objects, topology, proofs, and applications answer problems of perfectoid and $p$-adic geometry. “Uses an inherited architecture” is more precise than “fulfills Grothendieck’s vision.” 2121Scholze, “Étale Cohomology of Diamonds”.

Dessins provide a more direct programmatic route because the Esquisse became an organizing document for a named research community and a 1997 scholarly collection. Here influence is documented by publication, terminology, and explicit engagement. Even so, Belyi’s theorem precedes and enables the program, while later Grothendieck-Teichmüller theories contain constructions Grothendieck did not supply. A strong lineage is not a single-author lineage. 2222Grothendieck, “Esquisse d’un Programme,” 7–48; Grothendieck, “Esquisse d’un Programme,” 003; Schneps and Lochak, Geometric Galois Actions 1, 242.

The survival of questions

Legacy is often measured by answers: a theorem proved, a conjecture settled, a machine made to work. Grothendieck’s programs also survive as questions whose value does not depend on imminent solution: whether the absolute Galois group can be characterized through its action on profinite structures; which category of mixed motives has the expected properties; which models capture weak higher categories with usable coherence; and which form of tame topology retains geometric content without pathological excess.

An open question has several afterlives. It can remain active under its original wording. It can divide into precise subproblems. A counterexample can show that the proposed generality was wrong while revealing the right replacement. A new framework can make the old question ill-posed. Historians should record these transformations instead of treating “still open” as stasis.

The Esquisse is especially clear about provisional status. Grothendieck distinguishes tasks he regards as immediately approachable from characterizations he says may be out of reach and from principles he expects but has not proved. That grammar belongs in any account of his foresight. A conjecture expressed with confidence remains a conjecture. Later proof deserves later authorship; later failure belongs to the program’s history. 2323Grothendieck, “Esquisse d’un Programme,” 7–48; Grothendieck, “Esquisse d’un Programme,” 003.

The distinction prevents claims of lost archival gems from becoming an attribution machine. A useful find might be a powerful lemma. It might also be a question posed in a particularly fertile way, a counterexample never published, or a failed model that explains a later choice. Mathematical value takes several forms.

Teaching as transmission

Definitions survive because someone learns to use them. Grothendieck’s published work is famous for rebuilding foundations rather than treating them as preliminary. Later textbooks and seminars decide how much of that architecture a new mathematician must traverse. The Stacks Project says explicitly that it is not introductory; it addresses graduate students and researchers. Its hyperlinks allow a reader to move backward through dependencies rather than read thousands of pages linearly. 2424The Stacks Project Authors, The Stacks Project.

Legacy through correction

The same principle applies to historical claims made by Grothendieck. Récoltes et Semailles offers penetrating accounts of creation and severe allegations about credit. Later readers can value the former and dispute the latter. A mathematician’s authority in algebraic geometry does not settle an account of another person’s conduct. Conversely, a mistaken historical judgment does not invalidate a theorem. 2525Pierre Schapira, “A Truncated Manuscript,” Inference, December 2022, https://inference-review.com/article/a-truncated-manuscript.

Programs also need correction by plural communities. Anabelian geometry developed precise conjectures and theorems under later authors. Motive theories multiplied because one construction could not immediately meet every expectation. Higher categories admitted competing models whose equivalence had to be established. The variety is not a failure to inherit one true blueprint. It is how a broad vision becomes mathematics.

This view replaces the idea of fidelity with accountable transformation. A later mathematician need not imitate Grothendieck’s style, share his ethics, or accept his self-description to work on a question he posed. The relation can be critical. It can also be remote: a tool may pass through so many hands that direct knowledge of its origin is unnecessary.

What would count as a discovery in the boxes

If a researcher finds an unfamiliar argument in Montpellier, four steps separate finding from mathematical discovery. First comes identification: establish the exact leaves, sequence, date, and relation to known drafts. Second comes transcription: preserve notation and uncertainty, with another reader checking the result. Third comes verification: translate the argument into sufficiently explicit modern terms without silently repairing it. Fourth comes comparison: search published and unpublished literature for prior and later versions.

Only then can the result be classified. It may be a complete proof of a known statement, an independent proof, an incomplete anticipation, a counterexample, an exposition, or a new theorem. Each classification carries different credit. If modern mathematicians supply missing lemmas, their work belongs in the record. If notation hid an error, publishing the error with explanation can still serve history.

The process should culminate in a citable edition or article rather than an announcement about greatness. Other specialists need access to the transcription and images under appropriate terms. Claims of novelty should survive review. The archive’s prestige cannot substitute for those steps.

This standard may sound severe, but it protects the pleasure of a real discovery. When an archival idea withstands reconstruction and proof, its importance does not depend on mythology. It enters mathematics through the same public accountability expected of new work, while its historical form remains visible.

Later mathematics has already transformed many Grothendieckian questions. The boxes may alter chronology, reveal an earlier formulation, or recover an abandoned route. They do not control subsequent results. The living subject is the network of people who can test what the pages propose against independent mathematical standards today.

One final test concerns absence. Not every major modern field must be connected to Grothendieck, and not every use of categories, sheaves, or abstraction establishes a meaningful lineage. Probability, dynamics, analysis, geometry, computation, and mathematical physics have histories that intersect his work unevenly or not at all. A bounded account is more credible than a universal one. The map in this chapter selects paths for which texts and technical relations can be named. Its blank spaces are not deficiencies awaiting annexation; they preserve the possibility that mathematics develops from other problems, styles, and communities. Influence gains definition at its edge.

A living legacy is selective in both directions: later work chooses what it can use, and historical inquiry chooses which relations it can actually demonstrate. That restraint is historical accuracy. Each example makes that boundary concrete.