隐藏的工作坊

Between January and June 1981 Grothendieck wrote La Longue Marche à travers la théorie de Galois. The archive exhibition counts about 1,300 pages; Leila Schneps’s study says about 1,600 manuscript pages, with nearly as much again in addenda and developments. The difference likely reflects which sequence and supplements are counted. The defensible number is therefore a range, not a falsely exact total: roughly 1,300–1,600 pages in the principal manuscript, with extensive additional material. 11Grothendieck, La Longue Marche à travers la théorie de Galois; Maronne and Toën, Le temps des réflexions, 22, 33–34; Schneps and Lochak, Geometric Galois Actions 1, 59–66, 59–60.22Grothendieck, La Longue Marche à travers la théorie de Galois; Maronne and Toën, Le temps des réflexions, 22; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–60.

Six months and more than a thousand pages do not look like mathematical silence. They do look unlike publication. The manuscript was written for Grothendieck’s own use and not prepared as a finished book. Schneps warns that a later typeset transcription can mislead readers about the original form. The facsimile contains dense handwriting, marginal changes, internal questions, and conjectures revised as the work proceeds.33Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–61.

The title’s “long march” was accurate as working method. Grothendieck did not announce one theorem and spend the pages proving it. He explored a territory: moduli spaces of curves, their fundamental groups and groupoids, actions of the absolute Galois group, anabelian reconstruction, and a combinatorial description of arithmetic symmetry. Some proposals became weaker or changed form as counterpressure appeared. The document records movement.

From a drawing to a manuscript system

The Long March did not begin as a polished exposition of a settled system. It assembled questions about curves, moduli, fundamental groups, Galois actions, and combinatorial operations while revising strong early claims.

Schneps identifies moduli spaces, their groupoids, and compatibility across natural operations as a major manuscript theme. The archive exhibition places that theme after the late-1970s work.44Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66; Maronne and Toën, Le temps des réflexions, 22–31.

The mathematical scale remains visible without the tutorial. Grothendieck was no longer asking how one drawing behaved. He was asking how arithmetic symmetries could act compatibly across a family of spaces and operations. Compatibility, not a list of examples, made the project large. That proposition is supported by Schneps’s thematic analysis and by the manuscript’s extent; it is not presented as a theorem proved in six months.

What the sources say about low levels

The manuscript devoted extensive attention to low-genus and low-marking cases. Schneps’s table-of-contents analysis reports that later sections develop examples such as $(g,n)=(0,3)$ and $(1,1)$. This is a claim about documentary architecture. It does not, without page-level review, establish every formula or coherence relation previously explained in this chapter.55Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–60.

That separation also protects attribution. Later Grothendieck–Teichmüller formalisms cannot be inserted as stable definitions on every manuscript page. Schneps reports Drinfeld’s independent convergence.66Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 60–66.

The biographical fact is consequently stronger when narrower. A private manuscript of extraordinary extent worked systematically through low cases because Grothendieck regarded their relations as potentially controlling. The text preserved corrections instead of delivering a finished presentation. That working form, not a retroactive modern survey, is what made the workshop hidden.

The fundamental exact sequence

For a geometrically connected variety $X$ over a field $K$, with a chosen geometric point, the arithmetic étale fundamental group fits into an exact sequence

$$1\longrightarrow \pi_1^{\mathrm{\acute et}}(X_{\overline K}) \longrightarrow \pi_1^{\mathrm{\acute et}}(X) \longrightarrow G_K \longrightarrow 1.$$

The kernel is geometric: it classifies finite étale covers after extending scalars to an algebraic closure. The quotient $G_K=\operatorname{Gal}(K^s/K)$ is the absolute Galois group of a chosen separable closure. The extension records how field symmetries act on geometric covering data. Bresciani’s modern formulation treats arithmetic fundamental groups as extensions over $\Gamma_K$, while Schneps remains the historical witness for the 1981 manuscript’s program.77Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–61; Giulio Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” Algebraic Geometry 8 (2021): 231, 237, https://doi.org/10.14231/AG-2021-005.

All groups in this sentence carry profinite topology, and the homomorphisms used later are continuous. For a general reader, “profinite” means that the group is organized through its compatible finite quotients—the finite covering data—rather than as a discrete list of loops. That explanation has a limit: knowing all finite quotients is not the same as forgetting topology and keeping an abstract group. Continuity and the map to $G_K$ remain part of the arithmetic package.88Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–32.

The extension need not come with a splitting. If a section of the projection is chosen, conjugation by its image gives an action of $G_K$ on the geometric fundamental group. Changing the section by an element of the kernel changes that action by an inner automorphism, so the choice-free datum is an outer action. Bresciani uses exactly this conjugation relation when a section $s:G_K\to\pi_1(X)$ is present.99Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231, 237.

This is one point at which the source layers cannot simply be copied into a single sentence. Schneps’s 1997 overview calls the sequence split and writes a semidirect product. Bresciani’s modern account treats a section as additional data and explains its conjugation action; his quotient $\operatorname{Hom\!-\!ext}_{\Gamma_K}$ removes the corresponding target-kernel conjugacy. The canonical explanation therefore starts from the extension and introduces an action only after a section is chosen. Schneps remains evidence for how the manuscript was historically presented, while Bresciani controls the modern splitting and outer-action qualification.1010Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 60; Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231, 237.

The sequence is not peculiar to the Long March. It belongs to the étale fundamental-group theory developed earlier. The new question was reconstructive. How much of $X$ can be recovered from its arithmetic fundamental group and the associated Galois action?

A geometric base point is suppressed in much of the notation but not in the mathematics. Changing it identifies fundamental groups only up to conjugation, which is another reason inner automorphisms appear in the target of the hom conjecture. A groupoid can retain several base points at once, but no generic groupoid theory is required for this ladder: the necessary fact is simply that base-point choices prevent an unqualified equality of presented groups.1111Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–61; Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231.

For an elliptic curve or a higher-dimensional abelian variety, the fundamental group is highly abelian and cannot encode all geometry by itself. Grothendieck named the opposite regime anabelian—away from abelian. Hyperbolic curves, whose fundamental groups are strongly nonabelian, became basic candidates.

A punctured curve of genus $g$ with $n$ marked points is hyperbolic when

$$2g-2+n>0.$$

Examples include a projective line with at least three punctures, a once-punctured elliptic curve, and every smooth projective curve of genus at least two. Their nonabelian fundamental groups have enough structure to plausibly remember the curve.

The word “hyperbolic” carries a scope warning across the sources. The elementary inequality above covers punctured curves as well as proper ones. Bresciani’s introductory formulations of the full hom and section conjectures use smooth proper anabelian varieties over fields finitely generated over $\mathbf Q$, while his article separately explains conditional passages involving affine curves and orbicurves. The inequality identifies the intended nonabelian range; it does not erase properness, field, or marking hypotheses from a conjecture.1212Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–34.

The anabelian dream was that appropriate hyperbolic varieties over arithmetic fields could be reconstructed from their arithmetic fundamental groups, and that morphisms of varieties could be reconstructed from suitable continuous homomorphisms of those groups. Bresciani’s 2021 account emphasizes that the morphisms must lie over the same absolute Galois group and are taken modulo conjugation by the geometric kernel. The slogan “the group determines the variety” omits exactly the field relation and equivalence that make the conjecture meaningful.1313Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66; Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–32.

Why nonabelian information can remember shape

An abelianization records loops only after their order of traversal has been forgotten. On a punctured sphere, loops around different punctures satisfy one product relation, and the order and conjugacy classes matter. The full fundamental group retains how loops compose; its finite quotients retain this structure in profinite form.

Arithmetic adds another layer. Inertia subgroups associated with punctures or boundary divisors mark geometric locations inside the group. Galois acts on them in constrained ways. If a group is presented without labels, recovering which conjugacy classes represent inertia becomes part of reconstruction.

This explains Grothendieck’s attention to boundary structure. Compactifying a moduli space introduces boundary divisors corresponding to degenerate curves. Their incidence encodes how surfaces split and glue. Fundamental groups together with distinguished inertia around boundary components contain more geometry than an abstract profinite group alone.

The idea also explains tangential base points. A base point need not be an ordinary rational point inside a curve; it can be a tangent direction at a puncture, giving a controlled way to approach the boundary. Such points permit arithmetic fundamental groups and paths to be compared when ordinary rational points are unavailable.

These structures are technically demanding, but their underlying problem is clear. The network of internal symmetries and distinguished subgroups may retain enough information to identify the geometric object that produced it. Anabelian geometry asks for exact reconstruction theorems rather than resemblance.

The analogy stops where groups fail to retain all geometry. Not every variety is anabelian. Abelian varieties are an obvious warning. Even in hyperbolic cases, statements require arithmetic hypotheses. The Long March contains conjectural exploration, not a universal reconstruction theorem proved by Grothendieck in 1981.

Several meanings of “recover”

Anabelian statements come in levels. Object recognition asks whether two permitted varieties with equivalent arithmetic fundamental data must be isomorphic. Morphism reconstruction asks whether actual maps of varieties correspond to specified continuous homomorphisms of arithmetic fundamental groups, modulo the allowed inner conjugation. An explicit reconstruction procedure would ask for still more.

Bresciani makes the morphism level precise. For geometrically connected $T$ and $X$ over $k$, there is a natural map

$$\operatorname{Hom}_k(T,X)\longrightarrow \operatorname{Hom\!-\!ext}_{\Gamma_k}\bigl(\pi_1(T),\pi_1(X)\bigr).$$

Here the target consists of homomorphisms commuting with projection to $\Gamma_k$, modulo conjugation by the kernel on the target side. In the full hom conjecture, for $k$ finitely generated over $\mathbf Q$, suitable $T$, and smooth proper anabelian $X$, this map is conjectured to be a bijection. This is a claim about objects and morphisms over a field, not about an unlabeled profinite group.1414Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–32.

The forward arrow exists before the conjecture: a geometric map induces a homomorphism of arithmetic fundamental groups. What remains conjectural at full strength is whether the permitted group homomorphisms all arise geometrically and whether distinct geometric maps remain distinct after the stated quotient. In ordinary words, the invariant is functorial in the easy direction; reconstruction asks whether that passage loses nothing in the specified category. This distinction prevents “there is a natural map” from being reported as “the natural map is a bijection.”1515Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–32.

The group alone may not be enough unless one remembers the projection to $G_K$, inertia subgroups, decomposition groups, or other markings. “Isomorphic fundamental groups” must say whether the isomorphism lies over $G_K$. Omitting that phrase can turn a plausible arithmetic statement into a false abstract-group claim.

Rational points introduce a different reconstruction problem. A $K$-rational point gives, after a base-point choice, a section of the arithmetic exact sequence up to conjugacy. In Bresciani’s formulation, the section conjecture is the hom conjecture with $T=\operatorname{Spec}K$: for a smooth proper anabelian variety over a field finitely generated over $\mathbf Q$, rational points are conjectured to correspond bijectively to the permitted conjugacy classes of sections. The polished formulation belongs to Grothendieck’s later letter to Faltings and modern exposition, not indiscriminately to every 1981 page.1616Grothendieck, “Esquisse d’un Programme,” 7–48; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66; Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–32, 237.

A rational point therefore does not turn the extension into a canonically split product. It supplies a section after choices, and the relevant datum is its conjugacy class. Bresciani’s page 237 writes a section of the natural projection and the action it induces by conjugation on the geometric kernel. This is why the section conjecture belongs after, not before, the exact sequence and outer-action step: it asks which splittings have geometric origin rather than assuming that one preferred splitting was already built into the extension.1717Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 232, 237.

The distinction between geometric and arithmetic fundamental group is also essential. Over $\mathbf C$, two curves of the same topological type have isomorphic profinite geometric groups even when their complex structures differ. The arithmetic extension and Galois action supply the information that might distinguish them over number fields.

This is why the quotient $G_K$ cannot be dropped after the first displayed sequence. The geometric kernel sees covers after scalar extension; the arithmetic group records how those covers descend and are permuted over the original field. Object reconstruction and morphism reconstruction use the extension over $G_K$, not the kernel considered alone. Bresciani builds this requirement into the notation $\operatorname{Hom\!-\!ext}_{\Gamma_K}$.1818Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–32.

Thus anabelian geometry does not say topology alone determines algebraic geometry. It says nonabelian arithmetic topology, equipped with its field relation, can in favorable cases be rigid enough to reconstruct geometry.

This rigidity is most plausible for hyperbolic curves because their automorphism and covering theories are rich but constrained. It is least plausible for simply connected objects, whose fundamental group is trivial, or for highly abelian objects, where commutation erases structure.

Later theorems realized parts of the program under different hypotheses, but “parts” matters. Bresciani reports that Mochizuki proved a weaker hom statement for dominant morphisms of hyperbolic curves, corresponding on the group side to open homomorphisms. He explicitly warns that the full hom conjecture is strictly stronger because it concerns all homomorphisms. The section conjecture remains presented as a conjecture in that 2021 account. These status statements are modern verification, not evidence that every 1981 formulation was correct.1919Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–33.

The biographical importance lies in the reversal of viewpoint. Grothendieck’s earlier fundamental group classified covers of a known scheme. The new dream asked the classifier to remember the scheme itself. A tool became an archive of its object.

The seven steps should now be visible without collapsing them: a field gives $G_K$; a curve gives an arithmetic extension with a geometric kernel; a chosen section yields conjugation and the intrinsic outer action; hyperbolicity selects a nonabelian range; object recognition asks for isomorphism recovery; the hom conjecture asks for morphism recovery; and the section conjecture asks whether rational points account for sections. Exactness and induced actions are established structures. Reconstruction, full hom, and section statements remain conjectural at the generality just described. Schneps documents how the historical program developed; Bresciani supplies a modern statement and a 2021 status check.2020Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66; Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–33, 237.

Two major themes

Schneps groups the manuscript around two themes. The first is anabelian geometry: classify varieties determined by their étale fundamental groups and understand reconstruction over fields. The second is combinatorial Galois theory: characterize $G_{\mathbf Q}$ through its compatible action on the fundamental groups and groupoids of the moduli spaces $\mathcal M_{g,n}$.2121Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66.

The themes overlap but are not identical. The second organizes compatible moduli and groupoid operations across a system, whereas the first asks how arithmetic fundamental data can recover curves and maps.2222Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66.

Schneps identifies one especially strong early conjecture as untenable in that form, and the manuscript moved away from it. A question in §26 is therefore not silently treated as the position after §29.2323Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 61–66.

The manuscript’s self-correction, not a replay of the later tower tutorial, is the surviving claim in this chapter.

The manuscript as a record of correction

The archive exhibition reproduces a page on which Grothendieck states a principal conjecture and adds the French marginal note « C’est un peu faux, voir §9. » The object is a useful documentary anchor because it prevents a provisional note from being treated as a final statement.2424Maronne and Toën, Le temps des réflexions, 22.

Mathematical writing normally removes this kind of trace. A paper states the corrected theorem and perhaps thanks someone for a counterexample. A private log can preserve the false image, the point of failure, and the replacement. Grothendieck later made this visibility of error part of his ideal for “Mathematical Reflections.”

Preserved error has two historical values. It shows when a concept appeared, and it shows what resistance shaped it. It also creates citation risk. A reader selecting an arresting sentence from an early page can attribute a claim the author abandoned.

The marginal correction makes “Grothendieck’s conjecture” an incomplete description unless the section and date-state are specified. A question, a heuristic, an initial conjecture, and its later correction are different moments in the manuscript.

The manuscript’s form also affects page count. Rectos, versos, inserted addenda, successive developments, and later transcription do not map cleanly onto one printed pagination. The range in this chapter is not imprecision through neglect; it is a description of an archival object with layers.

Fifty-three sections and their afterlife

Schneps identifies §§1–53 as the main body. The first roughly 600 pages, §§1–37, were read and edited; the table of contents suggests that §§38–53 develop low cases such as $(g,n)=(0,3)$ and $(1,1)$. Addenda and further developments greatly increase the total.2525Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–60.

This architecture helps resolve page claims because a main sequence, addenda, and later transcription belong to different documentary layers. A page count must identify the counted unit rather than advertise the largest number.

The later typescript was useful precisely because handwriting restricted readership. But typesetting regularizes spacing, numbering, and symbols. It can make a private page resemble a submitted monograph and hide insertions or revisions visible in the facsimile. Critical use requires both.

Jean Malgoire’s role in transcription and custody belongs to the manuscript’s afterlife. It also continues the Montpellier student network. The same person who received the 1978 Belyi letter helped make the 1981 manuscript legible to later readers.

The text entered public mathematical history unevenly. Esquisse summarized themes in 1984. Specialized readers obtained transcriptions. Schneps’s 1990s study explained key conjectures. Full archival access arrived much later. Influence can therefore stem from summary, conversation, or independent rediscovery rather than direct reading of all fifty-three sections.

This staggered access limits influence claims. A 1991 paper resembling a Long March idea does not prove the author saw the manuscript; an explicit citation to Esquisse may transmit a theme without it. Similarity alone proves neither reading nor priority; known transmission routes require direct evidence.

Why he undertook the march

The archive exhibition says the anabelian dream began in 1977 in the wake of the surface work and that Grothendieck presented dessins ideas in 1978. Three years later, having found no student or colleague to undertake the exploratory voyage, he began writing the Long March himself. This explanation comes substantially from his retrospective account and should be attributed to him.2626Maronne and Toën, Le temps des réflexions, 22; Grothendieck, “Esquisse d’un Programme,” secs. 1–3.

His reported lack of a successor does not prove that every mathematician rejected a broadly circulated proposal. The program’s circulation was limited, its scope vast, and its form still changing. Some interlocutors may have pursued adjacent questions without accepting his total architecture. The sentence records his experience of lacking a relay.

Les Aumettes near Mormoiron is retained only as the address on later letters; the exact place of manuscript composition is not asserted.2727Maronne and Toën, Le temps des réflexions, 8–9; Grothendieck and Brown, Correspondance Grothendieck–Brown.

The archive exhibition uses an emphatic phrase for the pace of the 1981 interval. This chapter instead records dates and volume. Concentrated work supports no medical inference and needs no romantic explanation through deprivation or isolation.2828Maronne and Toën, Le temps des réflexions, 22.

A humane account can note ordinary material needs without inventing them. Grothendieck held a professor’s salary and had university duties. He lived at a documented address. Paper, copying, postal exchange, and time were the infrastructure. The hidden workshop was materially possible because academic employment continued even while conventional research outputs did not.

This fact complicates a heroic outside. The university he criticized also supported the life in which the manuscript was written. Public funds paid a professor whose most ambitious current work was not being published. In 1984 he would invoke responsible use of public funds when seeking a CNRS transfer.

The workshop thus stood both inside and outside the institution: economically and professionally inside, communicatively at its edge. That mixed position is more precise than exile.

Changing visibility

A layered manuscript requires a layered citation. The handwritten Long March, its addenda, a later typescript, the 1984 summary, and Schneps’s study are related objects, not interchangeable editions. A locator must identify which layer supports the claim.2929Grothendieck, La Longue Marche à travers la théorie de Galois; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–60.

Claim status changes within the manuscript as well. A conjecture on one page may be qualified in a margin or abandoned in a later section. The reproduced note that a statement is “a little false” demonstrates why title and page alone are insufficient. A citation must carry temporal status: proposed, corrected, retained, or later reformulated.3030Maronne and Toën, Le temps des réflexions, 22.

Typesetting changes appearance without changing authorship. Regular margins, continuous pagination, and normalized symbols can make exploratory writing look like a submitted monograph. The chapter therefore treats the transcription as an access aid and the facsimile as evidence of revision, insertion, and manuscript sequence.

Custody and editorial labor also belong in the citation history. Malgoire’s transcription work, Schneps’s analysis, and the University archive’s preservation made different parts of the object usable at different times. Naming those roles prevents “the manuscript became known” from erasing the people and institutions that created access.

Influence claims require one further step. A resemblance between a private page and a later paper is not proof of transmission. Schneps’s account of Drinfeld’s independent route is the clearest warning. The responsible comparison asks whether the later author read the manuscript, read the Esquisse, heard an oral report, or arrived independently.

The pages establish dated formulation and revision. They do not automatically establish public availability, uptake, or exclusive origin. Those distinctions are especially important because Grothendieck’s later moral writing made credit itself a central subject.

The manuscript’s internal references are another source of evidence. A direction such as “see §9” links a correction to a later state of the argument, but the link must be checked before the two passages are summarized together. Section numbering, insertions, and addenda can shift across facsimile and typescript. A citation that omits the version can make a correction point to the wrong textual object.

Translation adds a further layer. A French marginal phrase can be transcribed and located while an English rendering remains unapproved. The French wording and any English rendering are distinct textual objects; no draft translation is presented as quotation.

The chapter accordingly distinguishes access from endorsement. Digitization can make every page visible today, but it does not tell which pages circulated in 1981, which claims Grothendieck retained after revision, or which later mathematicians read them. An archive answers questions about survival and sequence; a correspondence answers questions about transmission; a publication answers questions about public form. Influence requires evidence connecting these layers rather than an assumption that present availability existed from the start.

The Long March was hidden from most mathematicians in the practical sense that it was unpublished, handwritten, and not broadly distributed. It was not hidden from its author, from every student, or forever. Jean Malgoire later edited and transcribed a substantial first portion. Grothendieck referred to the work in Esquisse. Schneps and others studied it in the 1990s. The University of Montpellier eventually digitized the archive.3131Université de Montpellier, “À la recherche d’Alexandre Grothendieck”; Université de Montpellier, “28 000 pages d’Alexandre Grothendieck”; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’”.

Visibility changed over time. In 1981 the text was a working manuscript. By 1984 some themes circulated through Esquisse. In 1995 an edited transcription of §§1–37 existed at Montpellier. Later online facsimiles and studies reached a wider audience. “Unpublished” is stable; “unknown” is date-dependent.

The archive exhibition gives the manuscript cote 140. That identifier matters because “Long March” can refer to the main sequence, addenda, or later transcriptions, each with a different pagination.

The same care applies to influence. Drinfeld developed a group later named Grothendieck–Teichmüller without having read the Long March or Esquisse, according to Schneps’s account. Resemblance can therefore demonstrate convergence and fertility without direct textual transmission.3232Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 60–66.

Calling every later related idea an anticipation by Grothendieck would repeat a cult of priority. The stronger historical statement is that the manuscript articulated a network of questions that later became active fields, sometimes through independent routes.

Hidden work and the problem of credit

A conventional paper fixes an author list, submission date, statements, and references. A private manuscript fixes some of these less securely. It can establish that an idea was written by a date, but not how broadly it was communicated or whether a later researcher could have read it.

The Long March therefore requires several distinct credit verbs. Grothendieck formulated questions in the manuscript. Malgoire edited a portion. Schneps analyzed it. Later mathematicians proved, reformulated, or independently developed related results. “Anticipated” is used only when the mathematical card compares exact statements.

This vocabulary matters because Grothendieck’s later writing made failures of attribution a major moral issue. A biography can report those accusations only after checking the cited works and communication routes. The existence of an earlier private page is relevant but does not alone prove appropriation.

Students face a particular risk in this history. An adviser’s boxes may contain broad suggestions surrounding a student’s theorem. The theorem still belongs to the person who proved and published it. Conversely, a student’s later paper should cite a specific communicated construction when evidence supports that relation. Neither rule is served by treating all ideas near Grothendieck as emanations of one mind.

Independent convergence, as in Schneps’s description of Drinfeld, is not an embarrassment to originality. Mathematical structures constrain good definitions. Two researchers can reach related equations because the same coherence problem demands them. Documentation should preserve both routes.

The hidden workshop is thus not a treasury from which later mathematics can be retrospectively claimed. It is a dated source whose contents must be compared claim by claim. That slower method is especially important for a biography committed to the ethics of credit.

A workshop with no public door

The hidden workshop had tools inherited from the earlier period: schemes, topoi, étale fundamental groups, descent, stacks. It had new raw materials: maps on surfaces, dessins, mapping class groups, and moduli of curves. Its product was not a theorem catalog but a research ecology.

An ecology contains failed branches. The very strong conjecture is one. It contains names not yet stable. “Anabelian” marked a contrast rather than a completed classification. “Teichmüller groupoid” could be formulated in several settings. The notes tested how the pieces fit.

The workshop also lacked an editorial division of labor. At the IHÉS, seminar talks could be written by participants, checked, reorganized, and published. Here one hand generated definitions, questions, corrections, and addenda. The enormous page count partly records the absence of compression by other readers.

Solitary authorship did not mean intellectual independence from others. Deligne, Belyi, Ladegaillerie, Mazur, Malgoire, Mumford, Thurston, and many earlier mathematicians entered the problems through theorems or conversations. The manuscript’s solitude was a production condition, not a claim that the mathematics had no genealogy.

The box as medium

By January 1984 Grothendieck would describe two voluminous boxes of handwritten notes accumulated during the preceding twelve years. The University archive later revealed the scale: thousands of mathematical pages among a much larger body. The box shaped what could be found, copied, and forgotten as well as storing paper.3333Grothendieck, “Esquisse d’un Programme,” sec. 1, p. 7; Université de Montpellier, “À la recherche d’Alexandre Grothendieck”; Université de Montpellier, “28 000 pages d’Alexandre Grothendieck”.

Pages in a box preserve sequence only if folders, numbering, and physical order survive. An author may know where a thought lies by memory; another reader sees handwriting without index. Grothendieck acknowledged that others might find the notes difficult to decipher.

The box allowed intellectual independence from publication schedules. It also increased dependence on the author for interpretation. A private notation could change without a glossary. A conjecture could be refuted fifty pages later. The archive’s later cataloging and digitization reduced physical inaccessibility but not mathematical difficulty.

The metaphor of hidden workshop should therefore remain material. Paper accumulated in identified places. Letters carried selected sheets out. Students and correspondents saw parts. Most of the profession did not.

The ethics of making hidden work visible

Grothendieck later issued restrictions on publication, and the archive’s legal and ethical history extends beyond this Part. Digitization establishes access, not authorization. Open online access is not blanket permission to reproduce manuscript pages; reuse requires an explicit statement from the holding institution.

Quotations are likewise limited. A private working note may now be public in an archive, but extensive reproduction can violate rights and overwhelm context. Short phrases with exact locators serve the biography better than decorative blocks of handwriting.

Mathematical review is especially important. The authority of the name can make a tentative note seem true. Every substantive statement drawn from the Long March receives a modern verification source or is labeled conjectural.

Privacy has a parallel rule. The mathematical correspondence contains occasional personal material. This chapter uses addresses, dates, and mathematical content needed for chronology. It does not extract intimate remarks because they happen to sit beside mathematics on a scanned page.

Afterlife without a prophecy story

The later importance of anabelian geometry makes it tempting to end the chapter with vindication. Major theorems did follow; dessins and Grothendieck–Teichmüller theory became active subjects. But vindication is too coarse for a manuscript containing revised and overly strong conjectures.

Some visions proved fertile as questions. Some precise forms changed. Some later definitions used tools unavailable in 1981. Some developments were independent. The afterlife is a branching tree, not the execution of a blueprint.

The manuscript’s greatest historical force may lie in conjunction: it placed curves, profinite groups, Galois actions, moduli towers, and boundary groupoids inside one research imagination. Later specialties often pursued one branch with far greater precision.

This division of labor is normal science. A program need not be completed by its author to matter, and it need not be right in every detail. The biography should neither dismiss the manuscript as speculation nor treat later theories as already complete within it.

The University archive now permits exact study of the earliest uses of terms, conjectures changed after counterexamples, and letters that transmitted definitions. Such work can replace legends with histories of concepts.

For Part IX, the relevant conclusion remains chronological. In 1981 a large private program existed. In 1982 an old higher-homotopy program reentered correspondence. Those facts form the route to 1983 without requiring the future to certify them.

The manuscript as an object

A page total is therefore a claim about a version. Rectos and versos, inserted leaves, numbered sequences, addenda, and later developments can enter or leave the count. A transcription introduces another pagination and can regularize a handwritten sequence without changing the words it transcribes. The chapter consequently retains the archive study’s lower estimate beside Schneps’s higher one and states separately that substantial supplements exist. This is not an evasive range; it identifies the present disagreement and refuses to compare unlike units. Any future exact number must state the archive cote, folders or section range, whether versos and inserts are counted, and whether addenda belong to the total.3434Grothendieck, La Longue Marche à travers la théorie de Galois; Maronne and Toën, Le temps des réflexions, 22; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–60.

The reproduced page 9 verso shows why material form affects mathematical status. A marginal correction points a reader from an initial formulation toward a later section. The page proves that revision occurred on the manuscript surface; it does not by itself identify the final version of the conjecture or authorize a direct English quotation. To cite the mathematics, an editor must follow the cross-reference, compare the relevant section, and identify whether a later addendum changes it again. The facsimile supports a claim about self-correction. A theorem-level claim requires the full textual chain and independent mathematical review.3535Maronne and Toën, Le temps des réflexions, 22; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66.

The document’s editorial state remained open. It was not prepared as a submitted monograph in 1981, and the later edited portion did not retroactively turn all fifty-three sections and their additions into a finished publication. Cote, section, manuscript page, transcription page, and printed discussion are consequently separate locator systems. This chapter uses the archive study and Schneps to establish architecture; it uses the reproduced leaf to establish visible correction; it uses later primary summaries only as retrospective testimony. An online image may improve access, but it does not collapse those witnesses into one edition or settle rights to reproduce the page.3636Grothendieck, La Longue Marche à travers la théorie de Galois; Maronne and Toën, Le temps des réflexions, 22, 33–34; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66.

Schneps’s 1997 essay is an informed witness to section architecture and the development of conjectures, not a substitute for every handwritten leaf. The exhibition facsimile is competent evidence for the visible marginal correction and archive context, not for the full surrounding argument. Primary page-level and mathematical review therefore remain open.3737Maronne and Toën, Le temps des réflexions, 33–34; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66.

Scale must be handled with the same discipline. Six months and a four-figure page range establish concentrated production; they do not establish a daily rate, continuous work on every day, or a psychological state. Addenda may have been composed alongside or after portions of the main sequence, and handwritten density varies. The proper comparison is documentary: sections, inserts, corrections, and later editorial access. Page count can show why transcription and navigation became difficult and why one summary could not replace the manuscript. It cannot measure mathematical quality by volume or convert intensity into diagnosis. This boundary is part of the object description, not an ethical aside added after the fact.3838Maronne and Toën, Le temps des réflexions, 22; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–60.

Four objects preserve different states. Cote 140 is the handwritten manuscript family; the 1995 Montpellier transcription covers the edited first portion; Esquisse is Grothendieck’s distributed 1984 summary; Schneps’s 1997 essay is a scholarly guide written after work on the transcription. A marginal insertion belongs to the first object, section organization may be reported from the editorial work, and a 1984 interpretation belongs to the later summary. None is a duplicate copy of the others.3939Schneps and Lochak, Geometric Galois Actions 1, 7–48, 7–48; Grothendieck, La Longue Marche à travers la théorie de Galois, 140; Maronne and Toën, Le temps des réflexions; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–61.

Schneps states her own competence and limit unusually clearly. Her note began as a brief, informal contribution to the Grothendieck day in Utrecht on April 12, 1996, and was meant to give only a partial idea of the manuscript. She had access to the edited first 600 manuscript pages, sections 1–37, and used the table of contents to describe sections 38–53. The difference between reading sections and inferring their focus from a contents list remains part of every architectural claim.4040Schneps, “Grothendieck’s ‘Long March Through Galois Theory’”.

The same essay explains why a clean transcription can misstate the feel of the object. Grothendieck wrote the pages for his own use and did not intend them for publication in that form; Schneps warns that the TeX presentation is misleading and directs readers back to the facsimile. In the mid-1990s she said the partial transcription could be obtained through Jean Malgoire. That is evidence of selective scholarly access, not broad 1981 circulation and not completion of the remaining sections.4141Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–61.

The Montpellier exhibition supplies the later public-access phase. It says that thousands of pages from the 1973–91 period were digitized and available online from 2017, while identifying the Long March as archive cote 140. Digitization changed who could inspect the surviving pages; it did not change their 1981 status, settle every internal revision, or show that a later mathematician had read them before the upload.4242Maronne and Toën, Le temps des réflexions, 22; Université de Montpellier, “À la recherche d’Alexandre Grothendieck,” no. 1.

Schneps also compares the manuscript with its first distributed synopsis. She describes Esquisse as a fifty-seven-page text written and distributed in 1984, summarizing some Long March themes in a more advanced form. She says that both works chiefly raise questions and examine approaches and contain few explicit theorems. A concise later synopsis can therefore witness public formulation without replacing the exploratory sequence, corrections, or page-level status of the 1981 object.4343Grothendieck, “Esquisse d’un Programme,” 7–48; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–61.

What reconstruction means

Take the projective line with $0$, $1$, and $\infty$ removed. With $g=0$ and $n=3$, the number $2g-2+n$ is $1$, so this single object lies in the hyperbolic range. Over a specified arithmetic field it has a geometric fundamental group, an arithmetic extension over $G_K$, and maps to or from other curves. The example lets three questions be distinguished without multiplying examples: whether the permitted arithmetic data recognize the curve; whether homomorphisms over $G_K$, modulo target-kernel conjugation, come from geometric morphisms; and, in the section specialization, whether rational points account for sections. Bresciani formulates the latter two as the hom and section conjectures and does not present either as a consequence of the inequality alone. For a general reader, the proposal resembles identifying an instrument from every compatible access test and from the way an external authority permutes those tests. The limit is decisive: finite étale covers are not physical tests, the arithmetic fundamental group is structured profinite data rather than an observation list, and no analogy proves reconstruction for this curve.4444Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66; Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–67, 231–33.

The failure boundary begins when information is discarded. A trivial fundamental group plainly cannot remember a rich variety by itself; abelianization deliberately forgets the order of loop composition; and even a nonabelian geometric group can omit the arithmetic field relation. Bresciani’s target $\operatorname{Hom\!-\!ext}_{\Gamma_K}$ retains projection to $\Gamma_K$ and then quotients by the specified inner conjugation, rather than using arbitrary abstract homomorphisms. A chosen section gives a conjugation action, but changing the choice alters it by an inner automorphism. Hyperbolicity therefore selects a plausible nonabelian class; it neither supplies a canonical splitting nor guarantees the full hom or section conjecture. In Bresciani’s 2021 status account, Mochizuki’s theorem covers the weaker dominant/open-homomorphism case for hyperbolic curves, while the full hom conjecture is expressly stronger. “Recover” thus always means recovery from a stated package under stated hypotheses, never all algebraic geometry from any group with a familiar name.4545Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 237.

Brown, Breen, and dated exchange

Five source fields define the first 1982 item in the edited correspondence. Ronald Brown is the sender; February 22 is the date; Professor A. Grothendieck at the Montpellier department is the addressee; Jack Duskin supplied the referral; and a possible stop connected with Brown’s June conference travel supplied the practical request. Papers were dispatched separately. This metadata shows how contact was initiated and why Brown wrote. It neither makes Duskin an authority on the contents of Grothendieck’s earlier program nor establishes that an in-person meeting followed.4646Wikiquote, “Alexandre Grothendieck,” 1982, 1–2; Centre for Grothendieckian Studies, “Correspondence”.

Grothendieck replied on March 4. Brown’s papers recalled a broad foundational synthesis, and Grothendieck referred to an extensive letter written to Larry Breen in 1975. He could not locate his own copy and wrote to Breen to request one. On March 25 he reported receiving from Breen an extensive reply and copies of three letters from the 1975 working program, then sent Brown the first. This sequence establishes retrieval through a correspondent: an old document did not emerge from a perfectly ordered personal archive but returned through the addressee’s retained copies.4747Grothendieck and Brown, Correspondance Grothendieck–Brown, 1982, 1–4; Centre for Grothendieckian Studies, “Correspondence”.

Brown’s April 15 reply explicitly acknowledges receiving the February 5, 1975 Breen letter and asks for the remaining letters. He describes points of contact with his own work and distinguishes strict models he regarded as developed from a weak higher structure he considered elusive. Grothendieck answered on May 5 and set out a broader local-to-global target. Receipt, evaluation, request, and response are separately dated. Brown was not simply handed a program; he compared it with results and limits from another research line.4848Grothendieck and Brown, Correspondance Grothendieck–Brown, 1982, 4–9.

The source is a typeset scholarly correspondence edition with French editorial apparatus, not proof that every original sheet survives in one folder. Page locators refer to that edition, rare omissions are editorially marked, and the cited claims are paraphrased. The chain is unusually strong because authors, dates, addressees, and acknowledgments interlock. It supports contact and document transfer, not an invented meeting, discipleship, or a private-state explanation for Grothendieck’s initial reluctance.4949Grothendieck and Brown, Correspondance Grothendieck–Brown, 1–9; Centre for Grothendieckian Studies, “Correspondence”.

The Breen documents also have multiple dates that must not collapse. February 5, 1975 is the date Brown gives for the first long letter he received. March 1982 dates Grothendieck’s unsuccessful search and Breen’s return of copies. April 15 dates Brown’s acknowledgment. May 5 dates Grothendieck’s response and promise concerning additional letters. Composition, custody, recovery, sending, receipt, and renewed discussion are six events. Calling all of them “the 1975 letter” hides the archival story by which the text became active again. A critical account should eventually identify each of the three recovered letters, surviving originals or copies, and their page ranges. The present edition supports the first chain; the complete stemma remains open.5050Grothendieck and Brown, Correspondance Grothendieck–Brown, 1–9.

The edited sequence also records asymmetry of expertise without settling priority. Brown stated the concrete results and limits of his program; Grothendieck supplied a wide topos-and-descent perspective and questioned its scope. Each criticized what the other lacked. Their letters are therefore evidence of a negotiation between research traditions, not a clean transfer of a preexisting solution. Where Brown names a theorem, proof credit remains with Brown and collaborators. Where Grothendieck names a target, it remains a proposal unless later work proves it. Where either man characterizes the other’s background, the statement remains correspondence testimony, not the narrator’s verdict.5151Grothendieck and Brown, Correspondance Grothendieck–Brown, 1–9.

No cadence claim is drawn from the surviving letters. Dates show intervals between outgoing and acknowledged mail, but postal transit, other correspondence, and missing items prevent those intervals from measuring attention or enthusiasm. The chain is used only where a letter explicitly acknowledges receipt, changes position, or requests another document.5252Grothendieck and Brown, Correspondance Grothendieck–Brown, 1–9; Centre for Grothendieckian Studies, “Correspondence”.

The edition itself is a defined witness. It was prepared by Matthias Künzer with Ronald Brown and Georges Maltsiniotis, uses French editorial apparatus around English correspondence, and marks rare suppressed passages. Its pages preserve the sequence of letters and editorial notes; they are not photographic proof that every original sheet survives together. This distinction matters because the chain can be strong at sender, date, acknowledgment, and mathematical topic while remaining open at physical copy history and rights.5353Grothendieck and Brown, Correspondance Grothendieck–Brown, 1–9; Centre for Grothendieckian Studies, “Correspondence”.

Brown’s March 11 reply makes the exchange less one-sided than a summary of Grothendieck’s program would suggest. He accepted the proposed Breen letter, mentioned prior exchange of offprints with Breen and a connection through Conduché, and explained why his group had concentrated on making a usable topological tool. He also named a concrete limitation: the methods then captured only part of homotopy theory. These are Brown’s statements of aim and boundary before he had read the 1975 letter, not retrospective commentary after agreement.5454Grothendieck and Brown, Correspondance Grothendieck–Brown, 2–3.

The May 24 letter provides a later receipt checkpoint beyond the chain ending on May 5. Brown thanked Grothendieck for the long letters and copies of the 1975 Breen correspondence, reported difficulty with the last copy both because it was hard to read and because the mathematics was difficult, and expected to discuss it with Breen at the Marseille conference. The physical legibility of a copy, mathematical intelligibility, and access to an interlocutor appear as three separate obstacles. This is direct evidence that transmission did not make a program immediately transparent.5555Grothendieck and Brown, Correspondance Grothendieck–Brown, 9–11; Centre for Grothendieckian Studies, “Correspondence”.

The sequence also revises the meaning of the declined visit. On March 25 Grothendieck thought a meeting would make little sense and redirected Brown toward Breen or Giraud. By April 15 Brown had read the first recovered letter and requested the others; by May 24 he had the copies and planned an in-person discussion with Breen. Contact did not end when the Montpellier visit was declined. It changed route—from a proposed meeting with Grothendieck to documents, replies, and another named interlocutor—without becoming a school or a transfer of authorship.5656Grothendieck and Brown, Correspondance Grothendieck–Brown, 3–11.5757Grothendieck and Brown, Correspondance Grothendieck–Brown, 1982, 1–9.

Brown’s August 17 letter adds an editorial state. Because the copy was difficult to follow, he prepared an informal English translation of one Breen letter, sent two copies, and asked Grothendieck to return one with amendments, additions, or corrections. The translation was therefore a working aid offered for authorial checking, not an approved substitute for the 1975 text. Brown also described several obstacles to developing the program and named Porter and Loday as active collaborators, widening the correspondence beyond a two-person exchange.5858Grothendieck and Brown, Correspondance Grothendieck–Brown, 14–16; Centre for Grothendieckian Studies, “Correspondence”.

On December 14 Brown enclosed an informal translation of the third Breen letter and said it might be useful to Grothendieck or his students. He also reported a result obtained with Loday and enclosed a research proposal for noncommutative homological algebra. The letter distinguishes resources already developed in Brown’s line from questions that might later reach topoi. Program recovery had thus become translation, comparison, collaboration, and a funding proposal by the end of 1982. None of those acts transferred Brown and Loday’s result to Grothendieck.5959Grothendieck and Brown, Correspondance Grothendieck–Brown, 16–17.

The translation sequence also places a firm limit on quotation. Brown’s renderings were informal and circulated for use and correction; the edited correspondence prints English letters but does not turn every working translation into an authorized edition of Grothendieck’s French. Historical claims may use the dated acts of translating, sending, and requesting corrections. Exact translated mathematical wording remains subject to comparison with the underlying Breen letters and human language review.6060Grothendieck and Brown, Correspondance Grothendieck–Brown, 14–17; Centre for Grothendieckian Studies, “Correspondence”.

Across ten months the document chain changed function. Brown’s February 22 referral opened contact; March letters negotiated a possible visit and recovered Breen’s copies; April and May letters compared programs and confirmed receipt; August introduced a working translation for correction; December supplied another translation, a new result report, and a research proposal. This is more than repeated correspondence but less than a jointly authored project. The dated sequence shows widening engagement, named collaborators, and changing document forms without converting postal intervals into a measure of commitment.6161Grothendieck and Brown, Correspondance Grothendieck–Brown, 1–17.

Program, conjecture, result

The Long March contains all three categories, and they must not be flattened. Program: classify anabelian varieties and investigate how arithmetic geometry might be reconstructed from nonabelian fundamental data. Conjecture: particular formulations propose that specified curves or compatible structures are determined in specified ways. Result: the manuscript also relies on established fundamental-group theory and analyzes cases, but its ambition is not itself a proved theorem. Schneps identifies an early “very strong conjecture” and then shows Grothendieck revising the route after judging the preceding approach too forceful. A later section therefore controls the status of the earlier proposal.6262Grothendieck, La Longue Marche à travers la théorie de Galois, 26–29; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 61–66.

Later status: Schneps’s essay, written for a 1996 meeting and published in 1997, is a historical witness to what specialists then understood of the manuscript and to the later results she named. Bresciani’s 2021 article is an independent modern source for the logical shape of the hom and section conjectures. The two sources are not stacked as interchangeable votes. Schneps supports the manuscript’s reception and historical development; Bresciani supports definitions, implications, and a dated modern status statement.6363Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66; Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–33.

Bresciani’s introduction identifies two forms of Grothendieck’s main anabelian conjecture. The hom form asks for a bijection between geometric morphisms and homomorphisms of arithmetic fundamental-group extensions over $\Gamma_k$, modulo target-kernel conjugation. The section form is its specialization to $T=\operatorname{Spec}k$, turning geometric morphisms into rational points and group homomorphisms into sections. His article studies implications between strengthened forms; it is not a proof that all hyperbolic curves satisfy both conjectures.6464Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–33.

Bresciani’s Theorem C illustrates the difference between an implication and a solution. Under the appropriate full-faithfulness hypothesis it obtains full faithfulness for the natural map on $T$-points; under printability it obtains an equivalence. Those are conditional conclusions whose premises remain visible. They organize relations among conjectural formulations and do not supply an unconditional proof of the premise for every hyperbolic curve.6565Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 232–33.

This modern separation changes the historical judgment. Established fundamental-group structure predates the Long March; Grothendieck’s 1981 pages formulate and revise a reconstructive program; later theorems answer specified subproblems; and the general hom and section formulations retain different statuses. An early conjecture cannot be upgraded because a later theorem resembles one rung, and a modern statement cannot be projected back as the wording of every 1981 section. Schneps and Bresciani serve historical and mathematical roles, respectively, supporting fertility without global “vindication.”6666Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66; Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–33.

Bresciani’s own results illustrate why present status must be read theorem by theorem. Theorem A concerns when his strengthened “printability” and full-faithfulness properties are geometric. Theorem B compares hyperbolic orbicurves with hyperbolic curves. Theorem C derives a hom-type conclusion from printability or full faithfulness, and Theorem D extends conditional statements to elementary anabelian stacks. None is worded as an unconditional proof of the general section conjecture for hyperbolic curves. The paper proves implications between formulations while leaving the central premise visible as conjectural.6767Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 232–34.

The date and genre of the modern witness also matter. Bresciani’s article was received July 4, 2019, accepted August 1, 2020, and published in volume 8 of Algebraic Geometry in 2021, pages 231–267. It reads Grothendieck’s later letter to Faltings and modern theorem literature; it is not a contemporary witness to what circulated in 1981. Schneps supplies that historical bridge. The two records support different halves of one sentence: what the Long March explored, and how selected conjectural forms were stated and assessed four decades later.6868Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66; Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures”.

Schneps describes her note as partial and informal; Bresciani works with modern conventions, including Deligne–Mumford stacks and the étale fundamental gerbe. His hypotheses remain attached to his theorem statements. Only the introductory group-extension formulation and status statements are used here, not every later generalization. The result is not a comprehensive 2021 survey of subsequent anabelian theorems.6969Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–61; Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–34.

Schneps’s account gives one internal correction a section-level chronology. In §26 Grothendieck posed a set of very strong questions; §§27–28 explored routes toward them; at the start of §29 he judged the preceding approach too brutal and changed direction toward smaller subgroups and additional structure. The historical result is not only that one conjecture was later false. The manuscript itself records a retreat from an overstrong route before the later external comparison. That is why an assertion from §26 cannot stand for the post-§29 program.7070Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 61–64.

The later Drinfeld comparison provides a separate credit boundary. Schneps states that Drinfeld had read neither the Long March nor Esquisse when he developed a strikingly related approach. Similar structures can therefore show convergence under a common mathematical problem without showing manuscript transmission. The 1981 date remains relevant to formulation, and Drinfeld’s route remains independently authored. Neither fact is improved by turning resemblance into appropriation.7171Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 64–66.

This internal and external evidence answers two different questions. Sections 26–29 show Grothendieck revising himself inside an unpublished manuscript. Schneps’s Drinfeld note shows later independent convergence outside its circulation. Bresciani then supplies a still later logical map of the anabelian conjectures. Revision, independent development, and modern implication are distinct historical relations; none turns the whole Long March into a proved blueprint.7272Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 61–66; Bresciani, “Some Implications Between Grothendieck’s Anabelian Conjectures,” 231–33.

From workshop to dossier

By the end of 1982 two documentary lines were present. One was the 1981 arithmetic manuscript, layered, corrected, and unpublished. The other was the Brown–Breen exchange, which recovered a 1975 letter and placed old foundational questions beside Brown’s active work. Their dates and sources should not be fused into one continuous book. The first has an archive cote, manuscript sections, and later editorial witnesses. The second has correspondents, acknowledgments, and an edited letter sequence.7373Grothendieck and Brown, Correspondance Grothendieck–Brown, 1–9; Grothendieck, La Longue Marche à travers la théorie de Galois; Maronne and Toën, Le temps des réflexions, 22; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–66.

The two forms also exposed work to different checks. Cote 140 preserves Grothendieck’s own corrections, but it does not by itself identify a contemporary external reader of every section. The Brown edition records dispatch, acknowledgment, comparison, and reply: Brown sent papers; Grothendieck recovered Breen’s letters; Brown confirmed receipt and asked for more; Grothendieck answered with another target. That sequence documents a program being tested in exchange without turning the correspondence into part of the arithmetic manuscript.7474Grothendieck and Brown, Correspondance Grothendieck–Brown, 1–9; Grothendieck, La Longue Marche à travers la théorie de Galois, 140.

By May 1982 the documentary contrast was concrete. The Long March remained one author’s vast handwritten exploration, known later through partial transcription and scholarly description. The recovered Breen material existed as copies sent to Brown, and Brown’s May 24 reply recorded both receipt and difficulty of use. “Hidden” therefore had degrees: one manuscript had limited editorial access; one older program moved through named correspondents; neither had the stable public form of a journal article. The evidence supports those routes, not a claim that the two programs fused.7575Grothendieck and Brown, Correspondance Grothendieck–Brown, 9–11; Grothendieck, La Longue Marche à travers la théorie de Galois, 140; Schneps, “Grothendieck’s ‘Long March Through Galois Theory’,” 59–61.

The threshold of 1983

The February 19, 1983, letter to Quillen opens the next dossier.7676Alexandre Grothendieck, À la Poursuite des Champs (Pursuing Stacks) (1983), Manuscript, https://webusers.imj-prg.fr/~leila.schneps/grothendieckcircle/PursuingStacksOrig.pdf, February 19, 1983.