Collective Intelligence

The credit line

Across EGA’s title pages, beneath Grothendieck’s name, appears a recurring formula: “with the collaboration of Jean Dieudonné.” Five words mark an arrangement too large for five words to describe. They identify an author and a collaborator, preserving an asymmetry. They also make it impossible to present the foundational treatise as the work of Grothendieck alone.11Grothendieck and Dieudonné, “Éléments de géométrie algébrique II: Étude globale élémentaire de quelques classes de morphismes”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique III: Étude cohomologique des faisceaux cohérents, première partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique III: Étude cohomologique des faisceaux cohérents, seconde partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, première partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, seconde partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, troisième partie”.

The line belongs to a series appearing in real time: EGA II and EGA III.1 in 1961, EGA III.2 in 1963, then the first three parts of EGA IV in 1964, 1965, and 1966. This was not a retrospective collected edition assigning an honorific credit after the work had become famous. Collaboration stood on the contemporary object by which the new geometry entered libraries.

Later accounts emphasize Dieudonné’s role in redaction, organization, and relentless standards of proof. The precise origin of every paragraph is not recoverable from a title page, and the biography should not pretend to partition the prose by temperament. The documentary truth is enough: Grothendieck’s conceptual program and Dieudonné’s collaboration produced a sustained text whose coherence mattered to the whole seminar enterprise.

To call Dieudonné a secretary would misunderstand mathematical redaction. A theorem changes when its hypotheses are corrected, its notation synchronized, its dependencies made explicit, or its proof fitted into a general proposition. The person performing that work helps determine, rather than only copy, the scope of the mathematical object that survives.

To call the pair equal coauthors would ignore the printed designation and Grothendieck’s intellectual direction. Accuracy lies between flatter legends. Collaboration can be indispensable without being symmetrical. A responsible biography does not solve the discomfort by moving all credit to one side.

The same rule governs SGA, where the printed roles are more numerous. Some volumes name co-directors; exposés bear individual authors; later redactors and editors assemble books. “Grothendieck’s seminar” correctly names his central direction and incorrectly assigns every line if used without qualification. The exact object under discussion must decide the wording.

The network was not one schoolroom

The SGA bibliography maps a distributed institution. SGA 3 names Michel Demazure and Grothendieck as directors. SGA 4 names Michael Artin, Grothendieck, and Jean-Louis Verdier. SGA 1 includes Michèle Raynaud in its book credit, while SGA 2 was augmented by an exposé by Michèle Raynaud. SGA 5’s eventual edition names Luc Illusie as editor.22Demazure and Grothendieck, Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3); Demazure and Grothendieck, Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3); Demazure and Grothendieck, Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3); Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3; Grothendieck, Revêtements étales et groupe fondamental (SGA 1); Grothendieck, Cohomologie locale des faisceaux cohérents et théorèmes de Lefschetz locaux et globaux (SGA 2); Illusie, Cohomologie ℓ-adique et fonctions L (SGA 5).

These credits identify different relations. Direction means responsibility for a program and its architecture. An exposé byline identifies responsibility for a particular treatment, not necessarily sole invention of every theorem cited. Editing may include reconstruction, harmonization, and correction years after the seminar. Collaboration on a title page can describe a sustained partnership. The roles overlap, but they are not synonyms.

The geography was distributed too. Artin’s Harvard notes on Grothendieck topologies came from a spring 1962 seminar. Hartshorne’s Residues and Duality, published in 1966, recorded a Harvard seminar of 1963–64 on Grothendieck’s work. Ideas traveled from IHES before revised SGA volumes existed and returned through mathematicians who had made them their own.33Artin, Grothendieck Topologies: Notes on a Seminar; Hartshorne, Residues and Duality: Lecture Notes of a Seminar on the Work of A. Grothendieck, Given at Harvard 1963/64.

Correspondence formed another route. The published letters between Grothendieck and Serre show sustained exchange across the years in which the program was forming. A letter could carry a question, example, conjecture, objection, or sketch faster than a fascicle. It also preserved address and date, features later treatises remove. The correspondence makes it impossible to portray Serre as merely one member of a school directed from Bures. He was an independent interlocutor whose arithmetic and geometric judgment repeatedly tested the program.44Grothendieck and Serre, Grothendieck–Serre Correspondence.

Tate supplied another external arithmetic pressure. His work on ℓ-adic modules and his 1965 conjecture on algebraic cycles and zeta functions spoke directly to the cohomological machine without making him an IHES disciple. David Mumford’s moduli and invariant-theory work, and contacts with American algebraic geometers, likewise belonged to a conversation wider than one institute. The network had a center; it did not have a closed border.55Tate, “Algebraic Cycles and Poles of Zeta Functions”.

IHES’s official retrospective presents Grothendieck’s seminar as central to its early identity. The institute supplied long stretches of time, visitors, and a freedom from routine teaching. Those conditions mattered. The myth of pure minds exchanging ideas outside material institutions obscures the salaries, travel, duplication, libraries, and editorial labor that allowed the network to function.66Institut des Hautes Études Scientifiques, “Alexander Grothendieck, Permanent Professor from 1958 to 1970”; Institut des Hautes Études Scientifiques, “The Foundation”.

Dieudonné: making a foundation citable

Grothendieck’s imagination often moved by enlarging the problem until a unifying structure became visible. Dieudonné’s mathematical character was equally powerful and differently directed. He had been a central member of Bourbaki, with its commitment to systematic exposition and exact foundations. Their EGA partnership joined conceptual expansion to a documented capacity for organized redaction.

The result included citability as well as clarity. A seminar expositor could refer to a numbered EGA proposition, know the conventions, and build a proof without rederiving the scheme language. A result about base change or finite presentation could be transported into SGA because its hypotheses had been stabilized. The reference architecture made distributed work possible.

Citability is a social technology. It permits one mathematician to rely on another’s proof without reproducing it and lets disagreements focus on a specific statement. Numbering, terminology, and indices seem secondary to discovery until dozens of arguments depend on them. EGA’s scale turned editorial consistency into mathematical infrastructure.

This infrastructure also disciplined Grothendieck. A sweeping idea entering EGA had to become definitions and propositions. Exceptions had to be handled; reductions could not remain private. Collaboration increased productivity and subjected the program to another mathematical judgment.

Their partnership should not be sentimentalized. The source base used here does not reconstruct daily negotiations or assign emotions to disagreements. What can be said is visible in the work: a sequence of fascicles carrying the same credit and a level of systematic detail that a single-author model does not explain.

The partnership also illuminates the eventual incompletion of EGA. A foundation of that ambition consumed time faster than a research program stood still. New derived and cohomological methods could require reorganization of earlier plans. SGA absorbed material that a projected EGA might once have contained. Collective production solved the labor problem only partly; the architecture continued to grow.

Serre: the necessary outside

Serre’s relation to Grothendieck preceded IHES and never fit a hierarchy of master and pupil. He had already transformed algebraic geometry through coherent sheaves, GAGA, and his work on algebraic groups and number theory. Grothendieck’s new framework absorbed and generalized many of these tools, while Serre continued to supply problems and judgments the framework had to answer.

The correspondence matters because the men’s styles were complementary without being opposites in a cartoon. Serre valued general theory but often pressed through examples and sharp questions. Grothendieck could pursue a structure further than an immediate application required. Their letters allowed a concept to meet resistance before publication.

An arithmetic example could expose a false general statement. A proposed cohomology theory could be measured against curves, Galois groups, or known comparison results. Conversely, Grothendieck’s categorical formulation could show that a phenomenon Serre recognized in one setting belonged to a wider construction. The exchange was productive because neither style simply ratified the other.

The letters are also a lesson in biographical evidence. A later memoir may compress years into a verdict. A dated letter places a question at a particular moment. It can show uncertainty, priority of communication, or the absence of a later term. But a correspondence edition remains selected and edited; silence in it does not prove ignorance. This chapter uses the letters to establish sustained interlocution, not to invent a complete private dialogue.

Serre supplied an outside that the program needed. He was close enough to understand its most abstract moves and independent enough to ask whether they paid mathematical rent. A research school made only of people trained inside one language could confuse fluency with truth. External mastery made the tests harder.

Artin: transmission and return

Michael Artin’s place in SGA 4 is fixed by the title-page credit as co-director. His 1962 Harvard notes offer an earlier documentary point. The theory of Grothendieck topologies was already sufficiently formed to teach, yet teaching it in another setting inevitably selected examples, order, and emphasis.

Transmission can be creative. To explain a new topology to an audience trained in algebraic geometry, one must decide which ordinary topological habits survive and which fail. Concrete sites and comparison problems test the axioms. A listener’s question can reveal that an apparently obvious functor lacks an adjoint or that a covering convention is ambiguous. The exported theory returns strengthened.

Artin’s later work on approximation and algebraic spaces would address existence and quotient problems at the edge of Grothendieck’s scheme category. Those results belong later; they should not be inserted as achievements of the 1962 notes. The historical continuity lies in the problems: representability, local structure, and the effectivity of geometric data.

In SGA 4, Artin’s concrete geometric force met Grothendieck’s topos architecture and Verdier’s derived methods. The triad should not be reduced to three personality types, but the mathematical roles were distinct. Co-direction meant that the finished theory had more than one center of competence.

The network’s international dimension mattered here. Harvard was an independent node. American algebraic geometry had its own traditions around Zariski, moduli, and arithmetic. Moving ideas across that context changed the audience that could criticize them. The “Grothendieck school” became influential partly by ceasing to be geographically containable.

Verdier: keeping all degrees together

Cohomology had produced sequences of derived functors long before SGA 4. Verdier’s developing derived-category language changed how those sequences could be organized. Complexes related by quasi-isomorphisms were treated as equivalent; a derived functor could be an object carrying all cohomological degrees and connecting maps at once.

For the six-operation architecture, this was more than elegance. Base-change maps, duality, tensor products, and direct images interact at the level of complexes. Working degree by degree can hide coherence. Derived categories make the compatibilities visible, though they introduce their own technical demands.

Verdier’s work grew in the environment of Grothendieck duality and the SGA programs. Hartshorne’s Harvard seminar volume records one transmission of the duality work. The mature published history extends beyond 1966. The chronologically safe claim is developmental: Verdier’s methods were becoming load-bearing while early EGA had been written without their later finished form.77Hartshorne, Residues and Duality: Lecture Notes of a Seminar on the Work of A. Grothendieck, Given at Harvard 1963/64.

This chronology shows how collective intelligence can force revision of an architecture. A new language could change the preferred form of earlier constructions rather than adding one chapter at the end. The prospect of rewriting a foundation grows as the foundation succeeds and attracts new tools.

Verdier’s name on SGA 4 protects against a common erasure in which “Grothendieck formalism” absorbs the derived category as one more invention of the director. The formalism is historical sediment: layers contributed by different mathematicians under the pressure of shared problems.

Demazure and the group-scheme program

SGA 3 named Michel Demazure and Grothendieck as directors and expanded across three published volumes. Group schemes required sustained expertise: finite groups, groups of multiplicative type, smooth groups, and reductive groups behaved differently over general bases. A single unifying definition opened the subject; it did not settle its structure theory.

Demazure’s role demonstrates that co-direction was one way Grothendieck multiplied a program without pretending to master every case alone. The relative viewpoint provided a framework. Detailed classification, root data, and degeneration demanded another mathematician’s sustained authorship and judgment.

Group schemes also linked the network’s themes. They were objects of descent; their torsors contributed to cohomology; their representations supplied coefficient sheaves; their bad fibers tested EGA’s morphism theory. SGA 3 was not a detour from the architecture of cohomology. It populated that architecture with symmetry.

The published date, 1970, comes after the seminar years 1962–64. As with SGA 4, the later volume records a mature object. Fine credit belongs to individual exposés and prior papers. “Demazure and Grothendieck directed” is accurate at volume scale; “they jointly proved every theorem” is not.

This granularity may sound cumbersome in narrative, but it restores the actual intelligence of the network. A general vision becomes a mathematical field only when specialists can work inside it, find its exceptions, and produce structure theorems the vision alone did not contain.

Two Raynauds, exact names

The documentary record demands a basic correction that broad histories often miss: Michèle Raynaud and Michel Raynaud were distinct mathematicians. SGA 1’s credit includes Michèle Raynaud; SGA 2 contains an added exposé by her. Michel Raynaud made major contributions to algebraic geometry and appears elsewhere in the network. An initial “M.” is not enough.

The danger is not a mere spelling error. When a woman’s work is folded into a better-known man’s name, collective history becomes less collective precisely at the point where its evidence is clearest. The remedy is ordinary scholarly care: read the title page, retain the given name, identify the role.

No further personal story should be invented to compensate. A biographer may be tempted to produce a scene of overlooked labor because the general pattern is plausible. Without a source, that would repeat the same erasure in sentimental form. The printed credits are strong, bounded evidence. They deserve to stand.

The Raynaud distinction also illustrates why bibliographic labor is intellectual labor. Mathematics cites by initials and surnames, often suppressing information needed by history. Recovering the collaboration network requires authority records, dates, and document-level bylines. A theorem’s proof and a person’s identity both depend on exact reference.

Illusie: a later recollection of entry

Luc Illusie later recalled beginning to attend Grothendieck’s IHES seminar in 1964. Identified accounts emphasize the energy, speed, clarity, and formal organization of the lectures. Because the testimony is retrospective, it should be presented as recollection, not used to reconstruct one unrecorded afternoon.88O’Connor and Robertson, “Alexander Grothendieck”; Schneps, Alexandre Grothendieck: A Mathematical Portrait; Artin et al., “Alexandre Grothendieck 1928–2014, Part 1”; Artin et al., “Alexandre Grothendieck 1928–2014, Part 2”.

The recollection is valuable because it describes apprenticeship from the receiving side. A young mathematician could enter without command of the whole subject and still perceive the formal structure. Clarity did not mean slowness or elementary content. It meant that dependencies were recalled and the architecture remained visible even as the exposition moved rapidly.

Illusie’s later editorship of SGA 5 is a different role and a different date. He helped turn the 1965–66 seminar into the 1977 volume. The interval separates his attendance from the later editorial role. Calling him “the editor” in a 1965 scene would backdate the book role; calling him only a student would erase what the volume later required.

His work on the cotangent complex and deformation theory would become another extension of Grothendieck’s architecture. The relative differential object had to remember singular and derived behavior that ordinary differentials missed. That later contribution extended the language rather than repeating a settled doctrine.99O’Connor and Robertson, “Alexander Grothendieck”; Schneps, Alexandre Grothendieck: A Mathematical Portrait; Artin et al., “Alexandre Grothendieck 1928–2014, Part 1”; Artin et al., “Alexandre Grothendieck 1928–2014, Part 2”.

Exposé, proof, theorem, name

Credit for a theorem in SGA begins with the exposé byline, followed by its introduction and citations. A result may be announced as Grothendieck’s, proved by the expositor from a sketch, adapted from another author, or strengthened during redaction. A later name attached by the community may compress this genealogy.

The safe narrative verbs differ. “Grothendieck proposed” describes a documented program. “The seminar developed” describes collective work when finer attribution is not established. “Artin’s exposé proved” is appropriate when the byline and text support it. “The published volume records” separates a 1970s redaction from a 1960s event. These verbs are not stylistic variation; they carry evidence.

Attribution is mathematically substantive because theorem statements evolve. A conjecture in correspondence can become a theorem under additional hypotheses in an exposé and a stronger theorem in the revised book. Assigning all three to the first person erases proof; assigning them all to the final expositor erases conception. The genealogy has branches.

Grothendieck’s wide influence magnifies the compression. Terms such as Grothendieck topology, Grothendieck topos, Grothendieck duality, and Grothendieck–Lefschetz formula can make one name sound like a complete author list. Eponyms are navigation devices, not source cards.

The same caution applies to “school.” A school can name a shared language and institutional center. It can also imply intellectual dependence. Serre, Tate, Artin, and Mumford had independent programs; Dieudonné had a mathematical career of enormous scope; younger participants developed theories that altered the framework. The network was denser and more reciprocal than a radial diagram suggests.

The original visual for this chapter therefore labels edges by documentary role and warns that it is not a map of friendship or influence magnitude. A line between two names can mean co-direction, redaction, correspondence, or an edited seminar. It cannot display the full human relation. The limitation belongs in the caption.

Authority and generosity

Later witnesses repeatedly praise Grothendieck’s generosity with ideas. He set out broad programs, told others about constructions, and allowed younger mathematicians to work in fertile territory. The resulting record supports the general picture of intellectual abundance.1010Jackson, “As If Summoned from the Void, I”; Jackson, “As If Summoned from the Void, II”; Cartier, “A Mad Day’s Work”; O’Connor and Robertson, “Alexander Grothendieck”; Scharlau, Anarchie; Schneps, Alexandre Grothendieck: A Mathematical Portrait; Artin et al., “Alexandre Grothendieck 1928–2014, Part 1”; Artin et al., “Alexandre Grothendieck 1928–2014, Part 2”.

Generosity, however, does not cancel authority. The person who defines the program can determine which questions count, assign exposés, and set standards of acceptable generality. A recipient of an idea may gain an important opportunity while also working in the shadow of its source. Both can be true.

Nor does later conflict retroactively falsify earlier collaboration. Grothendieck’s later Récoltes et Semailles judged parts of the mathematical community harshly and revisited questions of credit. That retrospective document belongs to the history of his memory and grievance, not as the sole authority on the 1960s. Contemporary title pages, correspondence, and bylines must remain primary.

The network’s productivity depended on trust: willingness to expose unfinished ideas, accept correction, and let another person write a treatment. It also depended on verification. Trust without proof would not produce SGA. The seminar’s collective intelligence arose from the alternation of openness and exacting check.

This combination helps explain the testimony about clarity. Grothendieck could share a vision because he had a language in which its parts could be assigned. Formal structure gave generosity a durable form. An idea offered only as personal insight might inspire; an idea accompanied by definitions and interfaces could become someone else’s research.

The youngest edge of the network

By the middle of the 1960s a younger generation was entering the architecture. Illusie’s recalled arrival is documented. Pierre Deligne, still very young, would soon become central to the Weil program and later to IHES. The exact date and content of every early encounter require careful sources; the later theorem must not be used to invent precocious scenes.

What can be said without invention is structural. The seminar had created entry points. A young mathematician could learn one exposé, master a functor, or work on a comparison problem, then discover that the architecture exposed new questions. Training occurred through responsibility for mathematics that mattered to the program.

The system thereby produced successors rather than replicas. Deligne’s 1974 proof of the Weil Riemann hypothesis introduced arguments not present in Grothendieck’s 1966 record, using a cohomological world built collectively. The distinction is a measure of the school’s success, not a diminishment of its founder.1111Deligne, “La conjecture de Weil I”.

Similarly, Illusie’s cotangent-complex work, Artin’s algebraic spaces and approximation, and Verdier’s derived methods extended the language where its original forms were insufficient. A living architecture invites alteration. If students only repeat the founder’s theorems, collective intelligence has stopped.

Grothendieck’s later sense of betrayal would complicate this story. In 1966 that later rupture was not yet a determined outcome. A chapter faithful to time cannot make every collaboration ominous because later relationships became strained. The record here is one of active shared work and asymmetrical possibility.

How a foundation passed between two mathematicians

No surviving title page describes the daily workflow between Grothendieck and Dieudonné, and this chapter does not invent one. The mathematical product nevertheless allows bounded inference. EGA sustains notation, numbering, cross-references, and a proof standard across fascicles separated by years. That consistency required repeated exchange between conceptual planning and detailed redaction.

Grothendieck’s program supplied the relative viewpoint and the demand for maximal functorial reach. Dieudonné brought deep mathematical judgment formed through Bourbaki and his own work across analysis and algebra. It would be crude to assign “ideas” to one and “writing” to the other. Redaction at this level chooses lemmas, orders dependencies, and sometimes reveals that an idea is false in its stated generality.

A hypothetical proposition about morphisms illustrates the process without claiming a historical scene. One first proposes that a property should be local on the base. To enter EGA, the claim needs a topology, finite-presentation hypotheses, behavior under composition and base change, and examples at the boundary. Each addition can change the proposition’s content. The final theorem is not the original sentence with polished French.

The credit line was therefore a compact governance rule. Readers knew whose program the book represented and that its realization involved another named mathematician. It did not expose contribution percentages. Modern demands for a contributorship statement cannot be retroactively satisfied, but the surviving formula should not be made less informative than it is.

The partnership also distributed cognitive load. One person could push the next research frontier while another maintained continuity with thousands of prior claims, though the roles surely overlapped. This division made speed possible without abandoning rigor. It also bound the program to the continued availability of both kinds of labor.

EGA’s eventual incompletion shows the limit. As derived categories, duality, and new moduli questions developed, a linear treatise threatened to require extensive revision. Even two exceptionally productive mathematicians could not freeze a field growing around them. SGA’s modular collective form became the more adaptable record.

The lesson is not that collaboration failed to complete EGA. It is that completion was a moving target produced by success. Every new theorem created foundational obligations. The pair had made algebraic geometry citable enough for many others to extend it faster than a final encyclopedia could absorb.

The life cycle of an exposé

An exposé began as responsibility for a segment of a program. Its author needed a statement, dependencies, and proof. The segment might originate in Grothendieck’s proposal, another mathematician’s theorem, correspondence, or the expositor’s own work. The byline identifies the treatment, while introductions and references reveal genealogy.

Oral presentation tested the segment against informed listeners. A missing hypothesis could surface. Notation could conflict with an earlier exposé. An example could show that a map was not canonical. The audience’s contribution rarely appears as a complete author list, which is one reason seminar history cannot be recovered solely from bylines.

A written fascicle then stabilized one version. Duplication allowed circulation, but copies might be scarce, corrected by hand, or superseded. A mathematician citing the fascicle participated in a provisional publication system. The document was more durable than speech and less fixed than the later book.

Redaction for a revised volume could reorder material, harmonize conventions, and incorporate improvements. The person doing that work might be an original expositor, a director, or a later editor. SGA 5’s 1977 credit to Illusie exemplifies the time depth: editing an eleven-year-old seminar was not equivalent to taking notes in 1966.

At each stage, “the same exposé” could differ. Mathematical identity was preserved through topic and argument, not exact textual sameness. A historian who quotes a 1977 sentence as spoken in 1966 confuses the descended object with one presentation. The correction is to cite edition and date together.

The life cycle made SGA unusually resilient. A theorem did not vanish when the meeting ended; a fascicle kept it active; a revised volume could become standard. It also made authorship layered. The durable version might owe its correctness to people not visible in the first announcement.

This is why claim-level source cards are indispensable. A volume-level citation can support the broad architecture. A priority or quotation claim needs the precise exposé and version. Collective intelligence is visible only when the evidence has enough resolution to separate its operations.

A grammar for disagreement

Shared language is often praised because it enables agreement. Its more important function may be to enable precise disagreement. If two mathematicians use the same definition of an étale morphism, they can isolate whether a proposed theorem fails at flatness, finite presentation, or a base-change step. Without common grammar, they may only report incompatible intuitions.

EGA and SGA supplied that grammar. Diagrams replaced ambiguous phrases about naturality. Universal properties replaced coordinate-dependent identification. Sites named what counted as a cover. Derived categories specified when complexes represented the same cohomological object. The language narrowed the place where a dispute could hide.

This precision did not make disputes impersonal. Authority, reputation, and institutional position still shaped whose objection received attention. A director could set the vocabulary in which alternatives had to be expressed. Formal equality of arrows does not imply social equality of speakers.

Yet the grammar created resources for resistance. A counterexample over a nonreduced base could defeat a famous conjecture. A missing commutative square could not be repaired by prestige. Once definitions were public, someone outside IHES could challenge a proof on its own terms.

Serre’s role illustrates the value of an independent fluent speaker. He could engage the general language without accepting every expansion of it. His examples were not external objections from someone who rejected abstraction. They were sentences in the same grammar that tested whether the larger claims were true.

Artin supplied another kind of disagreement through reconstruction. Teaching the ideas at Harvard required choosing what was essential. A different exposition can reveal redundancy or expose a silent dependency. Translation between mathematical contexts acts as peer review.

The common grammar therefore increased collective intelligence not by producing consensus but by making correction composable. One person’s counterexample could become another person’s added hypothesis; the revised theorem could then support a third person’s application. Disagreement entered the architecture as information.

People at different distances

The network contained permanent IHES professors, long-term visitors, Paris mathematicians, correspondents abroad, seminar expositors, and young attendees. These positions offered different access. Someone present for discussions could use an argument before it circulated. A correspondent might receive a focused account. A distant reader waited for a fascicle or publication.

Access affected contribution. To identify a gap early was to influence the theorem’s final form. To receive only the final book was to enter after choices had become invisible. The architecture aimed at universality, but its construction was geographically and socially concentrated.

Visitors partially opened the concentration. They carried notes and questions to Harvard and other centers, and brought outside problems back. International movement was not a decorative sign of prestige. It was a mechanism by which the theory encountered different mathematical cultures.

Language mattered. The core texts were in French, while international readers used English, German, Russian, Japanese, and other traditions of notation. An expositor working in French could still be translating concepts formed elsewhere. Later English textbooks would translate again, sometimes simplifying the historical credit with the terminology.

Translation can shift philosophy. The French revêtement carries the ordinary covering-space association; descente suggests return from above; site and topos acquire new technical lives. A literal English word may preserve the root and lose the contemporary resonance. This book keeps the French titles and supplies restrained translations rather than pretending equivalence is effortless.

Distance also protected independence. Mathematicians outside the immediate school could choose which parts to adopt. Dwork’s p-adic proof of rationality stood as a different route. Tate’s conjecture framed cycles through poles. Mumford developed moduli and invariant theory with his own priorities. The network was productive because it met alternatives, not because it monopolized the field.

The phrase “Grothendieck revolution” is therefore best understood as a change in shared conditions. One did not need to work at IHES or accept every abstraction to be affected. Definitions, theorems, and questions altered what counted as modern algebraic geometry across institutions.

Correction without humiliation

Research at the frontier makes error inevitable. A seminar culture can respond by hiding mistakes to protect authority or by treating correction as part of construction. The revised SGA volumes, with their changes and delayed redaction, show that correction was structurally expected, even if the social experience of particular corrections is not documented here.

Grothendieck’s pursuit of generality increased exposure to error. Each removed hypothesis opened new cases. A theorem known over a field could fail over an arbitrary base. Derived limits could introduce terms a formal analogy missed. Set-theoretic size could undermine an unqualified category. Ambition enlarged the surface that criticism had to inspect.

Collective work distributed inspection. Dieudonné could detect a foundational inconsistency; an expositor could find that a proof sketch required noetherian hypotheses; Artin could test a topos statement on concrete geometry; Serre could check an arithmetic consequence. These are representative roles grounded in the documented network, not claims about one particular undocumented exchange.

Correction becomes generative when the repaired statement has wider use. Adding constructibility to a sheaf theorem defines a robust coefficient category. Recognizing quotient failure motivates algebraic spaces. Controlling a derived limit clarifies ℓ-adic coefficients. The error locates the structure that was missing.

But a culture of strong leadership can make correction costly. Younger mathematicians may hesitate; a director’s conjecture can travel farther before being questioned. The record of productivity does not prove perfect psychological safety. Without testimony, this chapter neither idealizes nor invents conflict.

What can be assessed is the documentary outcome. The final corpus contains hypotheses, corrections, and multiple names. Its durability depends on having moved beyond one person’s uncheckable authority. Whatever the seminar’s interpersonal texture, the books submit claims to readers who were not there.

This submission is the ethical achievement of redaction. It replaces reliance on personal authority with an invitation to follow the maps. Mathematical objectivity is collective not because communities cannot be wrong, but because a claim is made available for correction by another community.

Michel Raynaud and robust geometry

Michel Raynaud belonged to the generation whose work made Grothendieck’s general language withstand difficult arithmetic cases. He must first be distinguished from Michèle Raynaud, whose SGA 1 and SGA 2 credits were discussed above. The shared surname is not a collective identity; every theorem and exposé requires its own byline.

Raynaud’s later celebrated results extend beyond the 1966 boundary, so this chapter does not use them to decorate his early presence. His importance to the network lies in the sustained algebraic geometry of group schemes, Picard functors, and specialization that the program required. These were areas where a general statement met bad reduction and nontrivial fibers.

“Bad” fibers were intellectually productive. Over a generic point a group or curve might be smooth and familiar; over a special point new components or infinitesimal structure could appear. A theory designed only for the generic object would discard arithmetic. Raynaud’s kind of geometry tested whether the relative language truly included degeneration.

The role differs from conceptual redaction. A difficult example can force a definition to change. A theorem proved in a singular or mixed-characteristic case can establish that a general framework has substance beyond formal extension. The network needed mathematicians able to live in those cases.

The title “collective intelligence” should therefore not imply that everyone contributed the same interchangeable reasoning. Diversity of mathematical resistance was the source of strength. One person generalized, another classified, another found a pathological fiber, another organized derived operations. The whole did not average these capacities; it composed them.

Nonabelian cohomology and Jean Giraud

Later accounts of the IHES circle identify Jean Giraud with the extension of torsor and nonabelian cohomology theory. The subject arose naturally because descent data often take values in automorphism groups that do not commute. Ordinary cochain groups no longer classify the situation.1212O’Connor and Robertson, “Alexander Grothendieck”; Schneps, Alexandre Grothendieck: A Mathematical Portrait.

A torsor under a group sheaf $G$ is locally isomorphic to $G$ with its translation action but may have no global section. Its transition functions form a nonabelian one-cocycle. Isomorphism classes are written $H^1(X,G)$, though this is generally a pointed set rather than a group. The loss of addition is mathematical information, not a defect of notation.

Gerbes move one level higher, organizing local objects that are themselves only locally nonempty and locally isomorphic. Their bands and degree-two classes require care beyond the early abelian formalism. Giraud’s later systematic work belongs to the development of this language; it should not be collapsed into one SGA 4 definition.

The biographical relevance is that Grothendieck’s architecture created questions whose solution required collaborators to broaden the architecture. A site and topos gave the field of locality. Nonabelian descent showed that set- and group-valued sheaves were not the end. The category of objects itself had to be allowed to vary.

This extension also changed the meaning of a moduli problem. If objects have automorphisms, a set of isomorphism classes loses stabilizers and may fail descent. A stack retains the groupoid. The seed lies in the problem; the mature solution belongs to a history with several authors and later dates.

Naming Giraud prevents a familiar retrospective error. Once stacks became known as part of “Grothendieck geometry,” contributions that developed their usable nonabelian theory could disappear into the adjective. A network history restores the people who made an opening inhabitable.

Pierre Cartier and the view from near the center

Pierre Cartier was close to Bourbaki and IHES and later wrote influential historical reflections on Grothendieck’s ideas. Such testimony occupies an intermediate position: it comes from an identified mathematician with firsthand proximity, yet it was composed retrospectively and shaped by knowledge of what followed.1313Cartier, “A Mad Day’s Work”.

Cartier’s account is valuable for concepts and atmosphere at program scale. It can describe the transformation of space, symmetry, and geometry across the period. It cannot be used without an exact locator to place one phrase in Grothendieck’s mouth or one person in a room on a specified date.

Witnesses also differ. One remembers generosity, another intensity, another the burden of redaction. Their accounts need not be reconciled into a single personality profile. People encounter a powerful director from different positions. Difference is evidence about the network.

Retrospective testimony tends to select turning points. Daily stretches of checking flatness or correcting references recede. The resulting portrait may be true and disproportionate. Documents restore the routine labor without denying the witness’s experience.

This book therefore uses identified recollection as one layer. It labels the witness and later date, checks factual anchors against primary records, and refuses composite scenes. The method may produce a less cinematic chapter, but it preserves the different viewpoints that collective history requires.

Hartshorne and the work of carrying

Robin Hartshorne’s Residues and Duality offers a concrete example of intellectual carrying. The title page identifies it as lecture notes from a Harvard seminar on Grothendieck’s work, held in 1963–64; Springer published the book in 1966. One object therefore records at least three roles: originating program, seminar transmission, and authored notes.1414Hartshorne, Residues and Duality: Lecture Notes of a Seminar on the Work of A. Grothendieck, Given at Harvard 1963/64.

Duality was technically demanding. Local residues, derived functors, and exceptional inverse images had to be assembled. Writing the notes required more than reporting that Grothendieck had a theory. Hartshorne made a version another audience could study and cite.

Calling the book “on the work of A. Grothendieck” credits the source. Naming Hartshorne as author credits the exposition and proof object. The formula is a model of layered attribution. Neither name makes the other unnecessary.

The Harvard location matters because it changed who could enter. A seminar in English-language American mathematics placed the duality program among students and researchers who might not attend IHES or read every French fascicle. The notes became a durable bridge.

Bridges can alter what they carry. A new exposition selects order and emphasis. It may supply proofs or examples absent from the originating sketches. Historical language should therefore say “Hartshorne’s account of Grothendieck duality,” not “Grothendieck’s book written down by Hartshorne.”

The same pattern recurs across modern mathematics. A theory becomes communal through texts written by people other than its originator. Their pedagogical and corrective work is one way knowledge ceases to be private property.

Deligne and the non-inevitability of succession

Pierre Deligne’s later place in the story is so large that hindsight pulls him into every mid-1960s scene. The source discipline adopted here resists that pull. Without a dated attendance record, the chapter does not assign him a chair in a particular SGA 4 meeting or invent an exchange with Grothendieck.

What is secure is the succession of mathematics. Deligne became Grothendieck’s student and later a permanent professor at IHES. His 1974 proof of the Riemann-hypothesis component of Weil’s conjectures used the étale-cohomological world while introducing arguments that were his own. The architecture enabled the proof; it did not predetermine it.1515Deligne, “La conjecture de Weil I”; Institut des Hautes Études Scientifiques, “Alexander Grothendieck, Permanent Professor from 1958 to 1970”.

This is a crucial model of collective intelligence across generations. A founder defines a field in which a younger mathematician can see a new route. The route may bypass the founder’s preferred conjectural bridge. Success then belongs both to the infrastructure and specifically to the new theorem’s author, in different senses.

Succession was not inevitable. A formalism can become sterile, or students can spend their energy defending it. Deligne’s achievement shows that Grothendieck’s language remained generative under independent pressure. It was strong enough to be used against the shape of its own open problem.

The later relationship between the men belongs to a more complicated history of credit, withdrawal, and memory. In this period it must not be foreshadowed as destiny. The younger generation was entering a flourishing program whose future was unknown.

The medal year intensifies the point. In 1966 the profession honored the founder while the person who would close the famous gap was not yet being honored for that future act. Chronology prevents prestige from flowing backward.

Work that leaves no theorem name

Seminar production included arranging schedules, duplicating pages, maintaining version lists, preparing bibliographies, correcting typography, and communicating with publishers. These tasks rarely carry theorem names. The present sources do not identify every person who performed them, so the chapter cannot supply a hidden staff by imagination.

Their anonymity in the record does not make the labor unreal. The survival of fascicles and the coherence of revised volumes demonstrate material work. A manuscript had to become an object that libraries could store and other mathematicians could read. Ink, paper, typing, and distribution were part of the knowledge system.

Mathematical notation made the task harder. Before modern digital typesetting, symbols could require handwritten insertion or special equipment. Errors in a subscript or arrow could change meaning. Copyediting technical French was not ordinary clerical transcription.

Rights and publication arrangements also shaped access. Later disputes about reprinting SGA show that ownership of the physical text was not identical with ownership of the mathematical ideas. A seminar conceived as shared research could become scarce through publishing history.

An illustrated biography should make this materiality visible through authentic title pages and fascicles when rights permit, not through invented photographs of people typing. The document is the scene that survives. Its worn or corrected surface can show labor without assigning a face falsely.

Recognizing unnamed labor has an evidentiary limit. One can state that production required it. One cannot attribute a particular task to a particular spouse, secretary, student, or staff member without a source. Ethical recovery is not licensed fabrication.

Names as infrastructure and hazard

Mathematics uses names to compress references: Grothendieck topology, Verdier duality, Artin approximation, Tate twist. An eponym helps speakers locate a cluster of ideas quickly. It is itself a piece of infrastructure.

The compression is hazardous. A named concept can have antecedents, collaborators, and later refinements. The person named may have proposed the framework while another proved the standard theorem. Different languages may attach different names. Repetition can make a historical convention appear to be a finding of fact.

Grothendieck’s name accumulated exceptional range. This reflects exceptional influence. It also creates gravitational force: nearby work is more likely to be described as his. The corrective is not to remove his name artificially but to restore bylines at the level where they exist.

Compound names can help and mislead. “Grothendieck–Lefschetz” places a new trace formula in an older genealogy but omits exposé authors. “Artin–Grothendieck–Verdier” accurately marks SGA 4’s direction and does not mean equal authorship of each page. Names indicate routes, not complete provenance.

This book’s fact cards therefore distinguish attribution from explanation. A modern reference may verify a theorem’s statement; only historical sources establish who did what when. A standard eponym is evidence of reception, not necessarily of origin.

The discipline is especially important in a life whose later writings contested credit. One should neither accept every later accusation as fact nor let standard names settle the matter. Documentary granularity is the alternative to partisanship.

The institute as a converter of time

IHES converted money and institutional autonomy into uninterrupted research time. Its founder Motchane, its small permanent faculty, and its visitors created an environment unlike a large teaching department. The publication and seminar records document sustained output during these years.1616Institut des Hautes Études Scientifiques, “Alexander Grothendieck, Permanent Professor from 1958 to 1970”; Institut des Hautes Études Scientifiques, “The Foundation”.

The conversion was not automatic. An institute can provide offices without creating a program. The seminar gave time a collective direction; EGA gave it a publication form. Visitors carried results outward, and outside questions returned. The institution’s success lay in becoming permeable to a network while maintaining a center.

Material support also produced vulnerability. A small institute depended on funding and directorial decisions. Grothendieck’s later break over military funding would show that institutional conditions were not morally neutral to him. But in 1961–66, no evidence permits writing the 1970 resignation as an intention already governing every seminar. The future conflict belongs to a later chapter.

What is visible now is the fit between institution and method. Grothendieck sought long chains of foundational work whose payoff might lie years ahead. IHES could support that horizon. The seminar divided the chain; Dieudonné’s collaboration stabilized it; international visitors multiplied it. Collective intelligence needed an institution capable of waiting.

Publication delays reveal a second timescale. The institute could host research before a commercial volume was ready. That gave insiders speed and created access inequality. Later editors paid the debt by producing books. The SGA corpus is the converted residue of institutional time.

A language no one person could exhaust

By 1966 the scheme language had become larger than any one of its makers. It included descent, representability, group schemes, topoi, étale and ℓ-adic cohomology, duality, and trace formulas. Grothendieck could direct their relation without personally exhausting every technical branch.

This is not a paradox. A language is successful when speakers can make true sentences its originator did not anticipate. The more functorial and interoperable the concepts, the greater this surplus. Group schemes could enter arithmetic classification; topoi could migrate toward logic; derived categories could reshape areas beyond duality. Later uses were evidence of portability, not proof of prior intention.

Collective intelligence names this surplus. It is not the sum of individual IQs or a claim that a group has one mind. It is the capacity of an organized network to produce, check, and combine results no member could hold in complete detail. Common notation, exact interfaces, and durable records made that capacity possible.

The network also generated error. Authority could propagate an overgeneral claim; scarce notes could conceal a gap; a revision could absorb credit; shared taste could neglect other methods. Collective work is not self-correcting by definition. Its strength depends on plural criticism and traceable documents.

Grothendieck’s achievement was therefore both personal and infrastructural. He identified wide-ranging connections and built conditions under which other mathematicians could act on them. The title pages show that they did. A biography that chooses either an isolated creator or an anonymous collective loses the architecture between.

Collaboration was not coauthorship everywhere

Modern readers often treat collaboration as a binary field in a database: either two people coauthored a paper or they did not. The Grothendieck network used more varied forms. A letter could shape a conjecture without producing a joint paper. A seminar expositor could write a proof inside a directed program. An editor could reconstruct a volume years later. A visitor could carry a theory between institutions.

These relations create different claims on credit. Joint authorship normally covers the published object as a whole. Acknowledgment may identify an idea or correction. Direction names architectural responsibility. An exposé byline identifies a local text. Historical prose should preserve the form instead of translating every relation into “worked with.”

The distinction also protects independence. Serre’s correspondence with Grothendieck does not make all their ideas joint. Artin’s co-direction of SGA 4 does not make his later approximation theorems products of the seminar. Illusie’s editorship of SGA 5 does not transfer every exposé to him. Relations connect bodies of work without merging them.

Conversely, lack of joint authorship does not imply isolation. EGA and SGA were ecosystems in which definitions, examples, and proof techniques circulated. A paper with one author could depend on years of collective testing. Formal bibliographic credit and intellectual context answer different questions.

The source cards therefore do not attempt a percentage allocation. They identify the documentary role and support each claim at that level. Where the evidence cannot separate conception, proof, and redaction, the prose names the seminar or volume rather than manufacturing precision.

Memory after the network changed

Most vivid accounts of Grothendieck’s seminar were written after the network had changed. Some witnesses had become established mathematicians; Grothendieck had left IHES and later withdrawn. Memory viewed the early period through knowledge of rupture and legacy.

This does not disqualify recollection. A witness can accurately remember speed, clarity, generosity, or pressure. Retrospection may even reveal patterns not visible at the time. It does require the biography to label who remembered, when, and from what relation to the subject.

Different witnesses can report apparently incompatible truths. A lecture can be formally clear to one listener and overwhelming to another. A director can be generous with ideas and exacting about the resulting work. Human descriptions do not obey a uniqueness theorem.

The source hierarchy of this account therefore places identified testimony below contemporaneous documents for exact dates and quotations, while retaining it for experience documents cannot supply. Title pages tell who was credited; witnesses tell how participation felt. Neither source can replace the other.

Grothendieck’s own later retrospective writing is one witness with unusual depth and unusual stakes. It records his memory and judgment. It cannot by itself adjudicate every earlier priority dispute. A collective history must not let the central subject become the sole court of his network.

Integration without homogenization

The mathematical aim of integration is to make results interoperable, not identical. A base-change theorem, a duality theorem, and a finiteness theorem retain different statements while fitting one trace formula. The seminar’s social integration worked best in the same way: contributors kept distinct expertise while their results met through common functors.

Homogenization would have been weaker. If every exposé copied Grothendieck’s voice without independent examples, the program could repeat one blind spot. Artin’s concrete geometry, Verdier’s derived methods, Serre’s arithmetic tests, and Dieudonné’s redaction supplied differences the architecture needed.

Integration required translation. A group-scheme result had to be stated in base-change language; a local calculation had to become a functorial map; a correspondence had to act on cohomology. The person performing the translation could discover that the proposed interface was wrong.

This is collective intelligence at its most exact: not many people agreeing, but many forms of expertise made composable. The architecture stores relations among results so no one person must recompute the whole.

The same principle guides this biography’s integration. Chapter prose links fact, mathematics, quotation, and image cards. It does not dissolve them into an uncited omniscient voice. A coherent life can preserve the provenance of its parts.

What the medal would hide

The approaching Fields Medal would condense this network into Grothendieck’s name. That condensation recognized his documented direction and mathematical innovations. It was insufficient as production history.

The official citation’s brevity could not explain how a general language was made reliable through other mathematicians’ work. It could not distinguish the 1963–64 SGA 4 seminar from its 1972–73 books or the 1965–66 SGA 5 seminar from Illusie’s 1977 edition. Ceremony has no room for three clocks.

Grothendieck did not attend the Moscow congress. His absence did not change the documented facts that he remained active at IHES and that the work being honored depended on collaborators with distinct credited roles across the seminar corpus.

The medal citation and the recipient’s absence must therefore be read against the mathematical shelf. The official record documented recognition and nonattendance; it did not supply a private motive or a complete account of collaboration.

Not consensus, not a group mind

Collective intelligence can sound like a claim that a community thinks with one consciousness. The seminar record supports nothing of the kind. Participants held different aims, levels of authority, and standards of useful generality. Their work became collective through interfaces and documents, not mental fusion.

Consensus was neither necessary nor sufficient. A theorem could enter SGA despite disagreement about its importance; an elegant program could attract consensus and still contain a gap. Proof, counterexample, and exact attribution remained the checks.

The network’s intelligence lay in memory distributed across texts and people. Dieudonné carried foundational continuity, Serre carried examples and arithmetic judgment, Artin and Verdier carried distinct formalisms, and younger mathematicians developed new branches. No inventory is complete, and no one node contained the whole.

Distributed memory becomes durable only when externalized. EGA numbering, SGA fascicles, letters, and revised volumes let a result survive turnover of participants. The archive was part of the cognition.

This view also leaves room for responsibility. A collective cannot be blamed or credited so vaguely that individual acts disappear. Grothendieck directed; named authors proved and wrote; editors revised; institutions funded. The network explains conditions, while source cards preserve agents.

The phrase is therefore descriptive, not celebratory. Collective intelligence can produce exclusion, hierarchy, and error alongside major mathematical results. Its achievement here was to make difference composable without making it disappear.

The plural subject

English grammar encourages a singular subject: Grothendieck invented, proved, created. Sometimes that subject is correct. At other times the documentary subject is plural: Artin, Grothendieck, and Verdier directed; Grothendieck and Dieudonné produced EGA under an asymmetric credit; the seminar developed; Illusie edited.

Choosing the plural is not an act of politeness. It changes causal explanation. The theory acquired its form because different competencies met, not because one finished mind was divided into clerical tasks. The printed names are evidence of that composition.

The singular remains indispensable for Grothendieck’s documented acts of conception and direction. Accurate biography alternates grammatical number according to sources. Its subject is one life inside a field that learned to think through many hands.

Provenance as social memory

Careful attribution remembers which document carries a claim, which person is named, which date belongs to a seminar, and which date belongs to a book. Without that memory, narrative fluency concentrates credit in the most familiar name.

Provenance also permits revision. If an original fascicle contradicts a later edition, the discrepancy can be located rather than hidden. If a byline corrects a secondary history, the prose can change without dismantling the chapter. Traceability makes collective history corrigible in the same way numbered propositions make a mathematical corpus corrigible.

The source system does not reproduce the whole social world. Conversations vanish, uncredited help remains hard to recover, and surviving archives reflect power. It is therefore a floor, not a claim to total knowledge. Its discipline lies in refusing to compensate for archival inequality with invented detail.

For a life so often narrated through superlatives, that refusal matters. Collective intelligence becomes visible through exact small records: a co-director on a title page, an exposé author, an editor named eleven years later, a letter with a date. The architecture of credit is built from these modest supports.

Every surviving credit therefore does double work: it anchors a mathematical object in time and preserves a human distinction inside the common language, allowing future readers to admire Grothendieck’s direction without mistaking direction for solitary manufacture of knowledge.

One chain, several documentary roles

The surviving record can place several roles in one evidentiary chain without turning them into one author. EGA provided published definitions of morphisms; Artin’s 1962 notes recorded a Harvard presentation of Grothendieck topologies; SGA 4 credited Artin, Grothendieck, and Verdier as directors.1717Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, première partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, seconde partie”; Grothendieck and Dieudonné, “Éléments de géométrie algébrique IV: Étude locale des schémas et des morphismes de schémas, troisième partie”; Artin, Grothendieck Topologies: Notes on a Seminar; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3.

The chain does not prove a simple flow from EGA to Harvard to SGA 4. Its dates show concurrent development and transmission. Establishing the direction of a particular idea would require letters, early fascicles, or exposé-level comparison. The chapter therefore distinguishes documented sequence from causal sequence.

Hartshorne’s 1966 Residues and Duality adds another role: authored exposition of a program identified on its title page as Grothendieck’s work and tied to earlier Harvard lectures. This is neither anonymous transcription nor independent origin.1818Hartshorne, Residues and Duality: Lecture Notes of a Seminar on the Work of A. Grothendieck, Given at Harvard 1963/64.

Illusie’s later editorship of SGA 5 adds temporal depth. The editor credit belongs to the 1977 volume, while the seminar title belongs to 1965–66. Combining those dates into one role would erase eleven years of mathematical and editorial development.1919Illusie, Cohomologie ℓ-adique et fonctions L (SGA 5).

This chain explains why credit must remain granular. Direction, exposition, transmission, and redaction can all be indispensable, but they answer different historical questions. The common language joined the work; the documents preserve its makers’ distinct positions.

Printed roles do not recover the whole relationship among collaborators. A co-director credit establishes responsibility for a volume; it does not measure friendship, frequency of discussion, or the origin of every idea. Those further claims require correspondence or identified testimony.2020Demazure and Grothendieck, Schémas en groupes I: Propriétés générales des schémas en groupes (SGA 3); Demazure and Grothendieck, Schémas en groupes II: Groupes de type multiplicatif, et structure des schémas en groupes généraux (SGA 3); Demazure and Grothendieck, Schémas en groupes III: Structure des schémas en groupes réductifs (SGA 3); Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 1; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 2; Artin et al., Théorie des topos et cohomologie étale des schémas (SGA 4), Tome 3; Grothendieck and Serre, Grothendieck–Serre Correspondence.

The limitation makes the title pages more, not less, useful. They provide a firm minimum that later narrative cannot erase. Where the record names direction, exposition, or editing, the prose keeps that role; where it remains silent, the chapter does not assign private labor by probability.

This evidentiary minimum also resists retrospective rank. A later famous theorem does not make its author the source of every earlier tool, and a director’s broad program does not make every exposé derivative. Dates and roles remain attached to the particular document in which they appear.

Recognition approaches

In 1966, the international profession prepared to award Grothendieck a Fields Medal. An award necessarily compresses. Its few lines emphasized geometry, K-theory, and the Tôhoku paper. No citation could list the exposé authors, editors, correspondents, and institutional conditions behind the cohomological program.2121International Mathematical Union, “Fields Medals 1966”.

Compression is not false if understood as a genre. Grothendieck’s individual contribution justified individual recognition. The error would be to let the medal rewrite the preceding six years as a solo performance. The documentary record is already more exact.

The award citation represented this work under one recipient’s name. Chapter 36 keeps that individual recognition separate from the documented production roles and from later interpretations of Grothendieck’s nonattendance in Moscow.

The documentary network remains visible behind the citation: Dieudonné’s credit line, Serre’s letters, Artin and Verdier’s direction, Demazure’s program, the two Raynauds’ distinct records, Hartshorne’s seminar volume, and Illusie’s later editorship. These records document how the mathematical corpus was produced.