Promises of help on the eve of Henri IV’s conversion
The reading, checks and editorial preparation used AI assistance.
Editorial correction · 5 October 2026. The earlier article (archived text) called this a “reproducible partial reading”. That overstated what readers could verify from the published page: the full working reading, coordinate ledger and verification package are not included. This revision describes selected passages and the research method, and makes that publication gap explicit. It adds no new decipherment evidence.
Paris, 22 July 1593. Henri IV’s conversion to Catholicism is three days away. The author of a letter attributed to Baudouin Desportes by the Bibliothèque nationale de France catalogue already sees his opponent’s intention to change faith drawing former enemies to his side. Continuing to resist on the strength of allies’ promises is becoming harder. To reject the prospect of peace, he writes, will take real action.
Much of this letter to Pietro Aldobrandini is encrypted. The working reconstruction presented here makes it possible to follow the author’s arguments on both sides of the leaf, although folds, areas obscured by ink and unidentified signs leave substantial gaps. It uses Satoshi Tomokiyo’s previously published key. This article presents selected passages from that partial reading and explains the method. The full working text and verification apparatus are not yet published here.
Prior reading and comparison update · 5 October 2026
Paolo Rosson has published a substantial reading of this same letter. His first recorded draft commit is dated 1 October 2026, before this article’s 2 October publication. Git timestamps record repository history; they do not independently establish when a text first became publicly available. Our earlier search missed this reading, so its result must no longer be understood as an absence of other decipherments.
Satoshi Tomokiyo’s follow-up, first posted 3 October and marked revised 4 October, credits Rosson and presents a partial image-overlay check. Tomokiyo states that he did not investigate alternative polyphonic assignments. This is a useful check of selected passages, not verification of every reading.
Our comparison began on 5 October. The texts agree over long passages but differ on consequential words and restorations. The article before this update (archived text) remains available. The original reading and its gaps have not been replaced by a composite text. A later check supports one proposed expansion, PAR, already present in Rosson; its scope and the remaining differences are recorded below.
Which letter was read
The source is held at the Bibliothèque nationale de France: Français 3984, leaf 186, recto and verso, images 347–348 in Gallica. The ending shows the signature Desportes, Paris and the date xxij Juillet 1593. Catalogue entry no. 88 supplies the author’s name, Baudouin Desportes, and the recipient, monsignor don Pietro Aldobrandini. The recipient’s personal name is not inferred from the courteous form of address on the leaf itself. Official BnF record · Recto · Verso
This matters for the scope of the study. The same collection contains other letters by Desportes, including a separate letter to Frachetta. Their contents cannot silently be added to leaf 186. Nor should the author be identified with the poet Philippe Desportes, or the catalogue’s Baudouin be replaced with Jacques: historical literature refers to several different Desportes.
Ordinary handwriting and cipher alternate here. The opening explains why the author has written infrequently: a truthful account might offend other participants, while bad news might distress the pope. Desportes then turns to complaints, explanations of earlier decisions and a defence of his own reputation. This progression helps explain the letter’s political meaning, but does not by itself allow unread passages to be filled in.
Why promises were no longer enough
One of the clearest passages is on the recto, in lines M3–M4 of the working edition. Joining the physical line break gives:
IL FAUDROIT BEAUCOUP D EFFECTZ AUANT QUE DE NOUS FAIRE REIECTER LA PARENCE D UNE PAIX
In English translation:
It would take a great deal of real action before making us reject the prospect of peace.
The French spelling follows the accepted reading. The expression about peace denotes its possibility or prospect; it does not say that peace has already been concluded. Spacing, punctuation and the English translation are editorial work. Manuscript, recto
In the neighbouring lines, the author reproaches allies for promises that have already deceived his side. Earlier, he describes an army that remained at the frontier and withdrew after a single rumour. The code identifying whose army it was remains unresolved, so naming it on the basis of historical expectations would be premature.
On the verso, the same argument returns in another form. V21–V22 expresses a willingness to accept a particular candidate as king, provided a way to protect him is shown. A lacuna remains between the parts of the sentence; the candidate’s name is unidentified. The repeated demand for effective protection, however, fits the contrast between promises and action on the recto. Manuscript, verso
This is the central theme of the readable passages: the conditions under which the struggle can continue and a political settlement can be accepted. References to the House of Bourbon and the kingdom’s fundamental laws in L10–L11 show that the dispute also concerns the succession. The gaps still prevent a reconstruction of all the proposals, letters and negotiations to which Desportes refers.
Three days before Saint-Denis
In B6–M2, the author writes about choosing a king at a moment when the opponent is beginning to attract former enemies through his intention to become Catholic and is due to attend Mass on Sunday. Henri IV renounced Protestantism at Saint-Denis on Sunday, 25 July 1593. The letter’s date and wording provide strong grounds for connecting this passage with him. This is a historical identification from context: Henri’s name has not been substituted for “the opponent” in the transcription. Letter · Château de Pau museum on the events of 1593
Seen in this light, the letter records the anticipation of change. The author understands that removing the religious obstacle may reduce the number of people willing to resist Henri. He is not yet describing a ceremony that has taken place, and does not know its consequences.
The letter also serves as a defence of the author’s conduct. Desportes recalls earlier warnings, explains his caution and anticipates criticism. His accusations show how he wanted to present the situation to Aldobrandini. Treating each of them as an independently established fact about his allies’ intentions would be a mistake.
How the reading works
Tomokiyo published the key to the Catholic League’s polyphonic cipher and described it through other documents. In this system, an ordinary sign permits two letters: for example, A/N, B/O or G/T. There are also separate signs for words, including QUE, QUI and POUR. Tomokiyo’s earlier study listed the encrypted parts of leaf 186 as unread; his October 2026 follow-up now discusses Rosson’s reading. A known key alone does not supply a finished text of this particular letter. Tomokiyo’s published key and study
The research records distinguish four layers:
- Image and sign class. The visible sign and its location on the original scan are recorded for each passage.
- Letter selection and word division. A letter consistent with French and the context is selected from the permitted alternatives. A coherent phrase remains a hypothesis when the sign’s appearance is uncertain.
- Ordinary handwriting. Unencrypted insertions are read separately. They cannot be presented as results of applying the key.
- Commentary and translation. Historical identifications and the modern translation are kept separate from the letter reading.
Several passages underwent separate repeat checks of sign classes without showing the checker the proposed phrase. One such check revealed a disagreement at M8:37: the sign may belong to F/S or G/T. Instead of the syntactically convenient PEUT, the edition leaves a gap. These AI-assisted checks apply to the specified passages; they are not external human expert review and do not certify every line of the letter.
Software checks whether the selected letters match the recorded classes. This does not establish that the signs were correctly identified in the first place, or that the resulting text is the only possible reading. No percentage of “decipherment accuracy” is calculated from the number of filled positions.
What a neighbouring document helped clarify
A primary control is particularly useful: a decipherment contemporary with the letters on leaf 185, and the corresponding encrypted text on leaf 188v. They allow individual signs to be compared with the words written beside them. This is a different document: its letter is dated 22 July, but has a postscript of 26 July. It cannot be used in its entirety as evidence of the situation on 22 July. Control decipherment, leaf 185 · Encrypted control, leaf 188v
The comparison provides strong support for two groups of signs in the letter under study:
- The Ω-like sign at B1:53 and V21:11 corresponds to the control sign annotated du Mayne. There are therefore strong grounds for relating the reference to patience in V21 to Mayenne.
- The g-like sign at B2_prefix:11 and B5:22 corresponds to the annotation legat, meaning “legate”.
The basis here is the shape of the signs and the annotated primary control. The full verbal form of the codes, articles and the preceding title sign remain separate questions. The identity of the nephew mentioned is also unresolved.
In the working transcription, these signs retain their earlier provisional labels; the proposed expansions belong to the commentary. X2, X3 and X4 are local to their respective blocks. The recurrence of a label elsewhere proves nothing by itself. In particular, not every X4 can be expanded as “legate”.
The tentative letter reading [MATEUCI?] also retains its question mark. For the historical narrative, it is enough for now to leave it as an unidentified name, without turning a plausible biography into evidence for the letters on the leaf.
What was already known
Tomokiyo’s key has been known since 2017; his original page records an update in 2023. The September 2026 cyphersolver attempt examined during preparation left this target unread. That earlier context is now supplemented by Rosson’s reading, recorded from 1 October, and Tomokiyo’s subsequent check. Neither this article’s search history nor the Git dates establishes global priority. Tomokiyo · cyphersolver materials
During the original preparation, a bounded search of selected historical editions found no printed decipherment of this exact leaf 186. This remains a record of that limited search, not a current claim that no other reading exists. The examination covered OCR texts and relevant passages in four editions: L’Épinois’s La Ligue et les papes (1886), volume V of Négociations diplomatiques de la France avec la Toscane (1875), volume I of Perrens’s L’église et l’état en France (1872), and Bernard’s Procès-verbaux des États généraux de 1593 (1842). Relevant pages in L’Épinois were also checked against the scans. OCR errors, variants of names, translations and publications outside this selection remain substantial limitations. This search establishes neither that the letter was unpublished nor global priority.
General criticism of Spanish policy and the legate has also long been known from neighbouring letters. On page 598, L’Épinois discussed it with references to leaves 184 and 177, rather than 186. The significance of this result therefore cannot be reduced to the discovery of previously unknown anti-Spanish sentiment. L’Épinois, checked page 598
The possible addition to knowledge lies in the precise wording of this letter to Aldobrandini, the order of its arguments, the conditions for agreement and the way the author defends his position. The selected passages provide a basis for discussion. A full assessment will require the unpublished working text and source apparatus alongside the images.
Comparing the readings
Rosson’s fuller text offers candidates for many gaps, including names and titles. Those suggestions need their own source checks. Agreement is not a measured accuracy rate: the editions divide signs and words differently, use different restorations, and share earlier source work.
Later source check: PAR. Rosson reads the paired-loop sign as par at the two places labelled V01:12 and V08:54 in our working edition. A separate AI-assisted check aligned the cipher on f. 176 with its contemporary decipherment on f. 177, without being given that proposed value. The control places the sign between the words corresponding to “comme” and “le passé”, supporting PAR there. The shared broad form supports PAR as a proposed expansion of the two target occurrences. This is a post-comparison addition, not a new discovery of the value or part of our original reading. The reduced image limits fine stroke comparison, and one control does not establish a rule for every similar sign. Pinned control image supplied by Rosson, from Gallica/BnF
Selected differences that remain unresolved
The identifiers below belong to our unpublished working edition; Rosson’s corresponding recto/verso line labels locate his passages in the pinned reading. They are not coordinates printed on the manuscript.
L02 / r36: property or lives
- Our preserved reading: BIENS
- Rosson’s reading: vies
- Unresolved point: Property versus lives is a sign/segmentation disagreement, not just spelling.
L15 / r49: a choice between verbs
- Our preserved reading: SONT
- Rosson’s reading: font
- Unresolved point: Both fit the recorded polyphonic classes; the word choice remains interpretive.
V17 / v16: one ream or two
- Our preserved reading: deux rames
- Rosson’s reading: une rame
- Unresolved point: Two reams versus one is an ordinary-handwriting question, independent of the cipher key.
V28 / v27: written letters and an editorial repair
- Our preserved reading: ME
- Rosson’s reading: [n]e
- Unresolved point: Rosson also reports ME as written, then emends it by sense. The written letters and editorial repair must remain distinct.
V30 / v29: a restored verb
- Our preserved reading: IOUEROIT
- Rosson’s reading: jour[ne]roit
- Unresolved point: A different verb and restoration; the versions cannot be silently merged.
These examples do not constitute a complete collation. No agreement percentage or blanket validation is claimed. Rosson’s repository describes AI-assisted production and no palaeographer review; Tomokiyo’s later partial check should be distinguished from that production process.
What readers can check here
The research materials prepared for this work include a full working French reading, a continuous Russian translation with commentary, and a package of verification materials. The package contains source images, passage coordinates, sign classes, accepted letter choices, a list of lacunae and records of separate repeat checks.
These supplementary materials are not included in this publication. Readers can inspect the linked manuscript and published key, and assess the selected French quotation and English interpretation above. The page does not provide everything needed to reproduce the full reading or rerun its verification checks.
Identifiers such as M3–M4 and V21–V22 refer to the unpublished working edition, not labels printed on Gallica. Without its coordinate ledger, they are not sufficient public locators for the passages. Full verification would require that mapping, the recorded sign classes and letter choices, and comparison of special signs with the neighbouring control document.
The package also contains a script that checks the compatibility of accepted letters with recorded classes and verifies file integrity through checksums. These checks passed when rerun during preparation of the article. The script checks the internal consistency of the records; palaeographic reading remains the researcher’s task.
A gap in the edition may mark an indistinct sign, an unknown code or an area obscured by a fold and ink. It does not always signify lost paper or a known number of missing letters. Words cannot therefore be completed across such places merely to produce smooth French prose.
What the letter contributes
In the readable passages, Desportes consistently demands grounds for trust: help, protection and sufficiently clear guarantees. At the same time, he tries to show that he foresaw the danger and acted reasonably. Three days before Henri IV’s conversion, this is a particularly telling witness to uncertainty within the Catholic camp.
The working reconstruction brings the specific words of this letter back into the discussion. Its value will depend on further checks: whether disputed signs can be clarified, connections between damaged sentences restored, and an earlier text of the document found. For now, the accurate description of the result is selected passages from a partial reading of a known letter using a published key, with lacunae preserved; the full verification apparatus is not yet included.
Sources and checked literature
- BnF, Département des Manuscrits, Français 3984, no. 88, f. 186r–v. Catalogue, 186r, 186v.
- BnF, Français 3984, control materials. Decipherment on f. 185, encrypted original on f. 188v.
- Satoshi Tomokiyo. A Polyphonic Substitution Cipher of the Catholic League (1592–1593). First published 14 May 2017; recorded update 29 June 2023. Article and key.
- dbourdeau. cyphersolver, materials on the sega1593 target, session of 17 September 2026. Research notes.
- Musée national du château de Pau. Materials on Henri IV and the path to civil peace. Historical account.
- Henri de L’Épinois. La Ligue et les papes. Paris: Victor Palmé, 1886. Digitised edition, scan of p. 598.
- Abel Desjardins, Giuseppe Canestrini. Négociations diplomatiques de la France avec la Toscane. Vol. V. Paris: Imprimerie nationale, 1875. Digitised edition.
- François-Tommy Perrens. L’église et l’état en France sous le règne de Henri IV et la régence de Marie de Médicis. T. I. Paris: A. Durand et Pedone-Lauriel, 1872. Digitised edition.
- Auguste Bernard. Procès-verbaux des États généraux de 1593. Paris: Imprimerie royale, 1842. Digitised edition.
- Paolo Rosson. Two cipher letters of July 1593. Repository, comparison snapshot c7d3688, first recorded draft. Consulted for this comparison on 5 October 2026.
- Satoshi Tomokiyo. An Intercepted Report of Henry IV’s Upcoming Attendance at Mass in a Polyphonic Cipher. Follow-up and partial check. First posted 3 October 2026; page marked modified 4 October 2026.
Abstract
This article presents a partial working reconstruction of Baudouin Desportes’s cipher letter to Pietro Aldobrandini, written in Paris on 22 July 1593 and preserved in BnF Français 3984, f. 186r–v. It applies Satoshi Tomokiyo’s previously published polyphonic key and distinguishes image-level transcription, contextual letter selection, ordinary handwriting, and historical interpretation. Contemporary controls on f. 185 and f. 188v strongly support specific signs referring to Mayenne and the legate; other names and substantial gaps remain unresolved. The readable passages stress the need for effective support and credible guarantees on the eve of Henri IV’s conversion. A 5 October update acknowledges Paolo Rosson’s reading, whose first recorded draft commit is dated 1 October 2026, and Satoshi Tomokiyo’s later partial check. Agreement, unresolved differences and a subsequent proposed PAR expansion are kept separate from the original edition. No new cipher key or global priority is claimed. The reconstruction and editorial preparation used AI assistance. This page publishes selected passages and source links; the full working text and verification apparatus are not included.