Every write-up

01 Write-ups

• read in 1931A Catalan letter in transposition, Barcelona 2 May [1413–16] — ACA Reserva 12resolved 18 Sept 20261413–16The oldest item on DECODE with a public image, listed as an undeciphered cipher of unknown type. It is a Catalan letter to the King written down the columns of each block, the word chunks cut by slashes, and Xavier de Salas explained the rule and printed the whole text in 1931, in the article the record itself cites. The rule is checked here on the image: the first block reads Senyor, jo us auie quezacom escrit… A royal financial officer in Barcelona, with Roger de Pallars, urges Ferdinand I to leave València and hold the corts, and defends his own accounts.
• adjudicated, not readThe Voynich manuscript — hoax, cipher or language, adjudicatedattempted 15 Sept 2026c.1420Six computational tests on the transliteration against eleven languages and implemented hoax generators, each re-run adversarially, and five literature sweeps. A plain or simply enciphered European language is excluded on transliteration-robust entropy; a verbose encoding and a structured meaningless text are left roughly even, with the tests that would separate them.
• readGaleotto da Ricasoli to the Dieci di Balía, Urbino 1425 — Readsolved 18 Sept 20261425A letter of Florence’s envoy to the Count of Urbino, listed on DECODE as non-decrypted. The abbé Gabbrielli glossed a few of its words in 1863 and built a key from them, which Meister printed in 1902; the letter was never printed. The three long runs he left unread are decoded sign by sign with his glosses as seeds, two of his nulls resolved and four new values confirmed on the siblings.
• read in 1869Alchymey teuczsch, 1426 — the sign alphabets of an alchemists’ notebook (Cod. Pal. germ. 597)resolved 18 Sept 20261426A German recipe book of a circle of gold-makers near Passau, dated 1426 and listed on DECODE as non-decrypted, with substance names hidden in three home-made sign alphabets. The compiler wrote the key on the first leaf as one unbroken run of signs “so that no one should understand that it is an alphabet”, then struck it through. Wilhelm Wattenbach read it in 1869 and printed three facsimiles with their plaintexts. Here the struck alphabet of f. 1r and the invocation alphabet of f. 6v are recovered from the images and checked against his readings, a third set is partly fixed from the Heidelberg catalogue’s reading of f. 93r, and the book is ruled out as a record candidate.
• not read, keys testedQueen María to Alfonso V, Valencia 1435 — Attemptedattempted 18 Sept 20261435The oldest unread ciphertext on DECODE with a public image: 42 signs inside a Catalan letter of Queen María to Alfonso V, written in Valencia on the day of Ponza, about something “continued until All Saints”, almost certainly the truce with Castile. Both printed Aragonese keys of the decade were tested: the 1437 alphabet is unrelated, and the 1429 nomenclator seems to match the word signs while its alphabet does not. Four neighbouring registers, 1,414 images, hold no cipher at all; the king’s own register shows his letter to her of 27 July 1435 was sent in ciffra.
• read in 1877Matthias Corvinus to Ercole I d’Este, Pozsony 1 June 1482 — DECODE R1156resolved 19 Sept 20261482The King of Hungary answers Ferrara’s call for help in the first weeks of the War of Ferrara, with troops, route, names and plan in a sign cipher. DECODE lists it as partly decrypted, but it was printed in 1877 and again by Fraknói in 1895. The runs are read here from the image and a working key rebuilt; four garbled places in the printed text are corrected, among them the closing sentence the edition prints as nonsense.
• read in partNicolò Sadoleto to Ercole I d’Este, Pozsony 1482 — the Venetian counter-offer to Matthias Corvinuspart read 19 Sept 20261482Ferrara’s envoy at the Hungarian court in the first summer of the War of Ferrara wrote parts of his despatches in a sign cipher; they were never printed, and DECODE lists four as partly decrypted. The alphabet is rebuilt from two sibling letters filed with their contemporary clear copies, which also shows that two of the four targets were read at the time. The seven cipher lines of 16 July 1482, which have no clear copy, are read in gist: the other side had offered Matthias Veglia, a fleet command for his natural son János Corvin and a hundred thousand a year. A faded block of 17 August gives fragments only.
• readMaffeo da Treviglio to Ludovico Sforza, Buda 1489–1490 — the key reconstructed in 2016 had never been turned back into signssolved 19 Sept 20261489–90Judit W. Somogyi reconstructed the cipher of Ludovico Sforza’s ambassador at the court of Matthias Corvinus in 2016, but printed it only as a table of numbers, with no image of a single sign, so nothing could be read with it. Her numbering turns out to follow the order of first appearance in the letter of 2 April 1490, which survives with its contemporary clear copy: matching the two gives the shapes back, and about twenty further signs came out of the reading. The two despatches that have no decipherment are now read. The four pages of 22 November 1489, of which only the first two lines had ever been deciphered, cover the peace with Frederick III, the Diet, the marriage of John Corvinus against the rival Neapolitan match, the queen’s “insatiable wrath of Juno”, and a warning that intercepted cipher letters had been read; the letter of 12 January 1490 reports the king riding to Vienna on the 8th and the ambassadors following on the 15th, which its own clear second page confirms word for word.
• keys found, part readEspagnol 318 — five ciphered letters of the Catholic Monarchs, and which keys open thempart read 17 Sept 20261497–1504The oldest item in the catalogue, and four different answers. No. 5 was never open: Ivan Parisi printed the whole text in 2020, having found that the king of Naples had his secret letter enciphered in the Spanish ambassador’s own cipher, by the ambassador’s hand. No. 92, the Great Captain to Lorenzo Suárez of 17 August 1500, is written in the Cifra general de los Reyes Católicos that Galende Díaz printed in 1994 — proposed by Tomokiyo from the look of the code groups, proved here against eight words a later hand wrote between the lines, with no misses; the key is transcribed, 683 groups, and the letter is readable. No. 95 has George Lasry’s alphabet and no published text: segmenting and clustering its 963 signs reads about two thirds, Spanish, about the French and about what a Señoría can be brought to do. Nos. 93 and 94 stay unread but are named: their code initials exclude both printed keys and point at the Gran cifra of 1501–04, whose only reconstruction sits in a Madrid manuscript that is not served online.
• 3 of 4 readFerdinand the Catholic to Jerónimo de Vich, 1511–12 — Ravenna, Milan and a spiritual war on Louis XIIsolved 18 Sept 20261511–12Three letters of Ferdinand the Catholic to his ambassador at Rome, AHN Estado 8715 N.45, N.57 and N.60 (April 1511 to September 1512), written wholly in cipher and filed without the decipherment their siblings carry. The nomenclator, a homophonic alphabet with 153 code groups for words and syllables, was rebuilt by lining two deciphered siblings up with the clerk’s text, and then read the three letters it was not built from. Julius II’s cardinals and Venice, Ferdinand’s account of how his army was pushed into Ravenna, the Sforza restoration in Milan. The 1515 letter is another system. Terrateig’s 1963 edition, not seen, may already print them.
• printed reading correctedAdrian of Utrecht, the Admiral and the Constable to Charles V, 30 December 1521 — a printed decipherment checked and correctedresolved 18 Sept 20261521AGS Estado leg. 8 no. 150, catalogued as “cifrada prácticamente en su totalidad”, was deciphered by Claudio Pérez Gredilla and printed by Danvila in 1899. The ciphertext was transcribed from PARES and aligned with his text on the group xif = V. M. That rebuilds a nomenclator of three-letter code groups and homophonic signs, shows that the dead king he called “de Inglaterra” is Manuel I of Portugal, and reads or corrects about seventeen of the groups he left unread. About ten are still open.
• read in partThe Duke of Sessa to Charles V, Rome, 18 April 1524 — a “1424” cipher that is a century youngerpart read 18 Sept 20261524DECODE catalogues RAH Salazar A-31 ff. 128–131 as a Non-decrypted letter of Luis Fernández from Rome, April 1424, and it was taken up as the longest ciphertext before 1450. The date is a typo: the letter is the Duke of Sessa’s despatch of 18 April 1524 and its duplicate, skipped by Bergenroth. Eight sibling letters in the same cipher carry the court’s decipherment; about 110 code groups and the alphabet were rebuilt from them by token tiling and template matching. The cipher now reads at the word level: the Pope’s secretary withholds the Bishop of Veroli’s letters, and Sessa asks the Pope for money for the Swiss. Four groups are attested nowhere and stay open.
• read in partRaince to Montmorency, Rome, 1526 — the key was misread by one columnpart read 17 Sept 20261526Eight ciphered despatches of the French embassy’s secretary at Rome sit in BnF fr. 2984; about 106 lines of them, the 13 May and 20 November 1526 letters to Montmorency, exist in no edition. The obstacle was never cryptanalytic. Tomokiyo published the key in 2020, but reading which glyph sits under which letter in his table by eye slips a column: l, m and n were each a place wrong, and every reading built on it was corrupt. Measured off the image instead — every ink blob within 15 px of a header column — the table resolves a control line glyph for glyph with nothing left over, and 86 of the 106 lines then read by hand from the microfilm. Nine days before the League of Cognac: the Castilians and the Bourguignons, le chemin de Valence, a capitulation for which ilz seront cause de la destruction, the plague in Rome, advices reaching the Imperials by their own people, and a man taken un jour, en plaine place pres du palais. Two months after the Colonna raid, a pontificate that seroit la cause de la totale ruine de sa maison. The clear close of 13 May, transcribed here, expects Andrea Doria at Civitavecchia within a day with six galleys.
• two of three readDel Vasto to Charles V and a letter to “Garbino”, 1527–28 — one found solved, one read, one without its keypart read 18 Sept 20261527–28Three imperial ciphers in BnF fr. 3022, listed with no decipherment. Del Vasto’s letter from Ischia of 27 September 1527 had been solved by George Lasry and Satoshi Tomokiyo. The anonymous “report” is del Vasto’s own letter of December 1527: Lasry had its letters, and about forty of its code groups are identified here, which makes it readable as his plan to march on Siena, Perugia and Urbino. The Madrid letter to “Garbino” is in Hieronimo Ranzo’s initial-letter code. Its numbering is not alphabetical, and without the table only its function words come out.
• key supplied, substance readSormano and Passano from Ferrara — eight failures, a published key, and two correctionspart read 18 Sept 20261529Three ciphered letters of François I’s agents at Ferrara, February 1529. Eight independent methods failed on a 1960s microfilm that will not separate thirty-six symbol forms — and then George Lasry’s 2023 key, published all along on a page I had not searched, read them. It also graded the attempt: 20 of the 22 values derived here were right, and the two wrong ones were exactly what had blocked it. The Duke’s answer comes out as “risolutamente concluse per cosa dil mondo non voler … se alcunamente accetare il regno” — Alfonso d’Este refusing the kingdom for nothing in the world. With the documentary findings from the failure: no. 63’s lost third sheet, the duplicate that cribs its twin, and Passano identified. The page also records two things this write-up previously got wrong. Contributed by Arya Sanketbhai Patel.
• readSormano and de Vaulx to François I, February 1529 — the duke of Ferrara declines the crown of Naplessolved 17 Sept 20261529Three ciphered despatches of the French agents at Ferrara, BnF fr. 3096 nos. 63, 65 and 66, listed unread beside two glossed siblings. Lasry’s key holds; the leaves add a null, a nomenclator for the duke and four letter forms. Eighteen thousand signs segmented from the Gallica scans, clustered and classified, then every line read on review sheets; the duplicate pair, which enciphers different stretches, checks itself. Alfonso d’Este will not take the kingdom or the captaincy of the French army, and the agents call his difficulties pretexts, eight months before Cambrai.
• readGramont, Mâcon and Langeac to Montmorency, 1529–1537 — four keys solved in 2023, four letters never readsolved 18 Sept 20261529–37Four ciphered letters to the grand maître from the French ambassadors at Rome and Venice, catalogued with no decipherment. Each has its own key, and Lasry had reconstructed all four from siblings; nobody had applied them. Read: Saint-Pol and the Venetians in January 1529; Clement VII on the council, his dream and his feigned illness on the road to Bologna in October 1529; the princes’ return in 1530; Paul III and the Farnese marriage in 1537.
• family identifiedThe Vatican cipher of April 1542 — an Antonio Elio cipherattempted 15 Sept 20261542Farnese to the nuncio Poggio, four folios, 6,549 digits, open since Lasry set it in 2019. Not read, but named: a polyphonic-syllabic cipher of the design Antonio Elio built for Paul III’s chancery. Seven sessions, five model classes excluded against matched controls, the Meister keys excluded under any renumbering, and the 400 dpi images read against the transcript: no separators, no decipherment, no key.
• solvedGarcía de Toledo to Philip II, 16 July 1565 — a relief of Malta waved off by the Grand Master’s firessolved 18 Sept 20261565The duplicate of the viceroy of Sicily’s ciphered despatch from Messina, two months into the Great Siege, AGS Estado leg. 1394 no. 247: clear Spanish with every sensitive word in two-digit figures, and no decipherment on the leaf or in PARES. Two runs beside clear words, seiscientos soldados and en tiera, gave eleven letters; the figures 12–43 run through the alphabet in order, so the rest was predicted and confirmed, and all 1,930 groups read. The Piccolo Soccorso got into the Borgo; a second run of galleys turned back four miles out on La Valette’s signals, as Mosiur de Lenni reported; and Sicily cannot pay for the war.
• readMary Queen of Scots to the Duke of Norfolk, “the 20th”, February or March 1570 — the one letter of the series never decipheredsolved 18 Sept 20261570BL Cotton MS Caligula C II f. 74r is the only letter of Mary’s ciphered correspondence with Norfolk in the Tower with no contemporary decipherment, dated only “the tventi of this instant”. Tomokiyo rebuilt the key from the three deciphered siblings and overlaid a partial letter-by-letter decode. Here the key is checked on f. 66r, and the letter is read in 23 lines from the new British Library IIIF images, stitched from tiles. Written weeks after the Regent Moray’s murder: Elizabeth blames Mary for the harquebus shot, Morton is rumoured on the move, and Norfolk is to write through the Bishop of Ross or Lady Scrope, never in his own hand.
• readMondoucet’s despatch of 13 July 1572 — a segmenter, not the archivesolved 17 Sept 20261571–74A Gallica sweep for volumes outside the standard lists found BnF fr. 16127, the Court’s file of Claude de Mondoucet’s correspondence from the Low Countries, 1571–74: about twenty ciphered despatches in one system, most with the Court’s decipherment, one long letter of 13 July 1572 without. Four aligners proved on controls had all sat at the shuffled baseline — not because of the archive but because an automatic glyph segmenter, the one stage never checked against a hand count, split looped d, crossbarred long-s and barred o/q into two components each and manufactured a 1.3–1.4 glyph-per-letter ratio. At glyph level it is an ordinary sixteenth-century homophonic substitution, one glyph per letter, four nulls. The key was built by hand from the 16 July crib in the same volume — the decipherer’s interlinear glosses on f. 62r and the verbatim reading at f. 64 — and a French 5-gram beam decoder then read the despatch. The decisive control is internal: decoding the 16 July block reproduces the decipherer’s own gloss verbatim. Partly enciphered, its clear bands give Spanish troop counts before Mons and German levies at Maastricht; the cipher carries Mondoucet’s reading of the Protestant position four days before Genlis’s column was destroyed at Saint-Ghislain and six weeks before St Bartholomew.
• readLanssac to Charles IX, Warsaw, 26 April 1573 — the Polish election embassy’s ciphersolved 17 Sept 20261573BnF fr. 4735 f. 124, catalogued “avec chiffre” with no decipherment. The key is a homophonic letter cipher with word signs, recovered from a sibling letter whose Court decipherment survives as gutter-cut marginal notes (“car je n’ay pas cinquante escuz”) and checked against the fragments on f. 124 itself. Both passages read, and the key then opens the three election letters whose only “decipherment” was that cut gloss: the Polish nation “autant vénale … comme sont les Allemans”, the Emperor’s three hundred thousand spent for nothing, and on 9 May the election carried against the Sultan, the Emperor, the princes of the Empire, Spain, Muscovy and Sweden, “qui tous estoient bandez contre vostre Majesté”. Tomokiyo’s published table for the cipher is corrected.
• solvedHenry of Navarre to Ségur — an alphabetical syllabary ciphersolved 16 Sept 20261585–86Three letters in figures to the envoy raising a German army, BnF 500 de Colbert 401 ff. 233, 239 and 288v, catalogued as undeciphered letters of Henry III. The upper figures pile into one slot in five (mod 5), the signature of a syllable table in alphabetical order; an annealer built on that structure, with the letters’ own clear French as context, returns a key whose letter part is alphabetical by itself and reads a matched control at 97 %. Conclude with Duke Casimir, raise the largest levy, march it at once; Clervant’s two thousand reiters; the Vivarais or Dauphiné road.
• decipherment foundAn anonymous figure cipher of October 1586 — the “unread” two thirds were deciphered in 1586, on the next leaf and between the linesresolved 18 Sept 20261586A League-period letter in two-digit figures, listed with a partial key and two thirds of its first page unread. The volume holds its own decipherment: folio 168 is a clear text of the cipher down to a cross mark, and from that cross to the end the decipherer wrote the plaintext between the lines. Tomokiyo’s key agrees with the glosses (Guienne, tres confident, Fumé), which also give 87 = e and the King of Navarre’s sign. The letter: Fumé, vice-admiral of Guyenne and Navarre’s Catholic confidant, keeps proposing a reconciliation between Navarre and the addressee, probably Guise, while Catherine de Médicis negotiates.
• resolvedPhilip II to Mendoza, 7 September 1589 — the “second undeciphered letter” is the decipherer’s copyresolved 17 Sept 20261589BnF fr. 3641 holds both of Philip II’s letters of that day twice: the originals in the general cipher Cg.13 and the 1589 decipherer’s fair copies, laid out like the originals with the unread code words in the margin, one of them miscatalogued as a second cipher letter. Aligning the pairs rebuilds some seventy syllables and fifty code groups of Cg.13, reads groups the decipherers left blank (respeto, vuestro, negocio) and corrects their “Francia” to England. Fourteen groups open.
• read in partA letter from Rome to the abbé d’Orbais, 23 August 1589 — the three keys bound with it do not fitpart read 18 Sept 20261589BnF fr. 3413 no. 62 was catalogued with three cipher keys bound in the same volume that might read it. None does. The letter, from a secretary of Cardinal Pellevé in Rome to Jean de Piles of the League’s council in Paris, is almost all in clear and carries the news of Henri III’s murder as it reached Rome. Its ninety-odd signs of cipher are in the Nevers–Piles alphabet that Tomokiyo identified. His partial table was filled out from a deciphered letter of 1586 in fr. 4715 in the same cipher. Read: Cassin, la protection, vostre regne, depeschera, mon maitre a esté retiré, and the signature, probably Baron. About thirty signs, most of them code signs, remain open.
• key recovered, read in partCharles III of Lorraine to Vaudémont, 18 June 1592 — a cipher catalogued as undecrypted, brokenpart read 18 Sept 20261592BnF Français 3621 no. 97, recorded in the DECODE database as Non-decrypted, with no key ever published and no crib in the volume: the January intercepts it was hoped to match exist only as a plaintext decipherment. An earlier attempt failed by treating it as a symbol cipher and clustering the glyphs by shape. They are ordinary cursive letterforms and can simply be read. The old solver was then shown, on controls with known keys, to be incapable of a cipher this size — 37–56% of letters at −2.64 where the true key scored −1.62 — and was replaced by a steepest-ascent search that recovers known keys at 98.7–99.6%. The key it found is verified against the manuscript, not the model: the group spelling chasteau occurs twice in the cipher, and Chasteauvillain stands in the clear on the same page. The Duke orders his son to conserve the plain and to bring the army back into the quarters of La Fauche.
• read, gaps markedPelissier to Jeannin, Burgos, 13 September 1592 — the League’s agent at Philip II’s courtsolved 18 Sept 20261592Nine pages, mostly in cipher, that Tomokiyo lists as only partially deciphered: the League’s agent in Spain reporting to Mayenne’s councillor four months before the Estates of 1593. Tomokiyo’s key for Pelissier’s later letters fits; calibrated on 1,764 signs of those letters aligned with their clear text, it reads 94 % of the ciphered words, every gap marked. Philip grants 500 ducats, Pelissier argues against holding the Estates now and for two armies at 300,000 écus a month, and reports the case being made in France for Navarre.
• 2 of 7 readNevers to Pisany, 1593 — the numerical key reads two letters, and the other five were never in itpart read 17 Sept 20261593Seven ciphered copies of the duc de Nevers’ letters from the road to Rome, catalogued as one office’s key. They are two. Tomokiyo’s Nevers cipher no. 46, read from the full-resolution table in fr. 3995 and checked against the office’s own decipherment of a Gondi letter, resolves every figure of the two letters to Pisany: the 8 September letter whole, the 14 October one in long stretches. The five letters to Revol are in the Court’s symbol cipher no. 60; one copy is in clear, one carries ninety symbols, three were not reached.
• two of three readHenri IV to Béthune, 9, 10 and 22 November 1601 — two letters read, and the ceiling on the third measuredpart read 17 Sept 20261601Three ciphered letters of the King to his ambassador in Rome, BnF fr. 3484, catalogued “avec chiffre” with no decipherment. The notice’s “copie du n° précédent” turns out to be the minute of the 10 November letter in clear, and aligning it with the cipher recovers a Villeroy-office key of the design Bazeries printed for Béthune’s brother in 1599: ten thousand écus to Cardinal Aldobrandini for the Peace of Lyon. The 22 November letter is now read too — it interleaves clear French with its cipher, and carries the Camaiano pension, Barberini’s coming and the dauphin’s baptism. The 9 November letter, a solid page of cipher, stops at about six words in ten, and an oracle bound shows why: the decoder is already past the ceiling its transcription allows.
• key and text foundHenri IV to Savary de Brèves, 1603 — the key in fr. 3462 and the letters in print since 1853resolved 18 Sept 20261603Twelve royal letters of 1603 to the ambassador at the Porte, BnF fr. 3541, “avec chiffre” and listed as undeciphered, with a partial reading by Tomokiyo from a rebuilt key. The office’s own table survives as fr. 3462 f. 103, and Berger de Xivrey printed the letters in 1853 from a contemporary copy in clear: three in full, nine as summaries. Tomokiyo’s overlay of f. 60 is checked against the print (revolteur d’Asie, protestans). The cipher leaves are not digitised, so the letters have not been read against the key.
• readHenri IV to Landgrave Maurice — the figures Rommel printed in 1840 and the key he printed in 1846solved 17 Sept 20261602–09Rommel set the King’s ciphered passages to the Landgrave of Hesse-Kassel as rows of figures he could not read, then printed the key six years later with one sample and stopped. Put together, they read all seven passages, about 4,100 groups: Bouillon and the German princes, the Gunpowder Plot and Mérargues “forgés sur mesme enclume”, two million livres for the Dutch, and on 20 May 1606 the call to the princes to take counsel together against a Spanish King of the Romans, “pour la conservation de la liberté germanique”. Contributed by Arya Sanketbhai Patel.
• resolvedHerbault to Béthune, 13 February 1626 — the decipherment three folios awaysolved 18 Sept 20261626The one letter marked “avec chiffre” and not “avec chiffre et déchiffrement” among some thirty of 1625–26 to the ambassador in Rome. It is not an undeciphered text: no. 26 is not a second letter but the same despatch, carrying the same cipher with the plaintext written between the lines in 1626. Identity fixed on the word-for-word clear text, the cipher runs falling at the same points and matching groups. The ciphered passages are the papal Legate and “le Pape pour l’acheminement de ces troupes in la Valteline”, the admission that France made peace with the Huguenots because “le faix d’une double guerre ne se pouvoit plus supporter”, the Dutch squadron sailing home with the Huguenot admiral, and Savoy’s attempt to engage France against Spain six weeks before Monzón. Contributed by Arya Sanketbhai Patel.
• solvedFrom Warsaw, 24 December 1627 — an alphabet in plain ordersolved 16 Sept 20261627DECODE R1408, one page of figures and stray letters with only a dateline in clear, listed as non-decrypted. The letters a and m are nulls, the letter pairs are alternates for consonants, the rare groups are syllables, and the alphabet, recovered blind by a 5-gram-plus-dictionary annealer against shuffled controls, turns out to run in order: odd figures a to m, even figures n to z. A reminder that a promised canonry of Olmütz for one of the Queen of Poland’s sons has not been conferred.
• solvedRichelieu to M. de Rancé — BnF Français 3829solved 14 Sept 20261629Ciphertext-only recovery of a homophonic alphabet with a doubling mark and nomenclature; then found to agree word for word with the decipherment Avenel printed in 1858, which the catalogues had missed.
• solvedMaltravers to Ormonde — a regular block ciphersolved 15 Sept 20261634–35Doubled letters written with consecutive figures betray a regular key (consonants three figures each from 7, vowels from 64, nulls 91–111). Every spelled word reads, and the nomenclator is then confirmed clause for clause against Wentworth’s own dispatches in Knowler’s Strafforde’s Letters (1739): the King refusing Kildare, and Ormonde moved for the Council “in exchange” for Sir Piers Crosby.
• read in substanceCharles I and Nicholas to Boswell — a regular key, four signs and the wrong addresseesolved 15 Sept 20261643Two ciphered letters of 2 November 1643 (TNA SP 84/157). The alphabet is Robert Pitt’s: a 24-letter row four times over 20–115, verified here at z = 9.6 against 20,000 permuted rows. Added here: the four graphic signs are word-signs introduced inside the spelled word (good, Cousin, Master, us), and the King’s letter is his re-credentials to the Duke of Courland’s envoy, sent through Boswell in Boswell’s cipher. Both letters read in substance.
• solvedFra Giovanni di Lucca to the Emperor — a polyphonic figure ciphersolved 16 Sept 20261644DECODE R2159, 231 dot-delimited figures of Italian, listed as non-decrypted and left half-read on Tomokiyo’s page because its crib is self-contradictory under any substitution. It is a 24-figure alphabet in which 17 stands for both i and n and 19 for both t and s. Eight crib letters fixed, the rest annealed, every seed agrees, shuffled controls do not; a Viterbi pass over the two alternatives reads it end to end: an offer to keep the Turk off Rákóczi, raise Moldavia against him and give two thousand Cossacks.
• read by others, verifiedSir Richard Forster, 13 May 1644 — verified blind, and why the first attack failsresolved 15 Sept 20261644Already read by Lasry, Biermann and Pitt; Tomokiyo’s page still says unsolved. Not a regular Stuart key but a mixed homophonic alphabet of 34 symbols over 207 tokens. Pitt’s key stands at z = 8.8 against 20,000 permutations, and an annealer recovers 31 of 34 symbols blind — but only after the French model writes u for v and i for j; before that it reads five of six matched controls and still fails the target.
• readThe prince de Conti’s ciphered mémoires, March 1649 — the Fronde’s instructions to its envoys with the Archdukesolved 18 Sept 20261649BnF fr. 3854 nos. 41–43. The two “undeciphered” mémoires of 26–27 March were read by Lasry in 2023; they are read again here from the leaves, with three small corrections. The third, whose decipherment was meant to be the crib, is a different number code, and its clerk never glossed the first twelve lines. A key rebuilt from about 3,300 glossed groups reads them: Conti to Laigue, 5 March 1649, explaining the Rueil conference away to Archduke Leopold Wilhelm.
• octastich readUrquhart’s Cyphral Octastich — a book cipher on his own Jewel, read without the plaintextsolved 16 Sept 20261652Eight lines and a “Decagram” of numbers on the last leaf of The Jewel (1652), Schmeh’s Top 50 no. 28. Number k indexes a word on physical page k of the book, and Urquhart took the first word of the initial he needed: a habit checkable without any plaintext. The public transcription was thirteen numbers short; the 1983 edition’s photographs on the HCPortal give 285, which decode straight with 238 of 284 first-occurrence hits against 0.43 for shuffled and random-page controls, into a royalist prayer for Charles II. Vals AI published the rule in August 2026 as a Claude Fable 5.1 result; the companion distich claim does not reproduce.
• readBordeaux to Brienne, London, 30 May 1653 — read with the English key sheet, once the transcription was rightsolved 18 Sept 20261653Thurloe’s intercept of the French envoy’s despatch, BL Add MS 4200 f. 88: 810 tokens of a Brienne-office cipher whose syllables run alphabetically through three marked series. A solver built on that design read its controls but not the letter. The Deciphering Branch’s own sheet for “Mr. Bordeaux”, Add MS 32263 f. 1 (DECODE R7537), is the key; checked against the page images, the transcription had merged two signs (ꝺ = u, ∂ = n), dropped tokens and lost marks. The Council heard the deputies of rebel Bordeaux before Cromwell gave the King’s envoy his audience, and Brienne is told to hold back French mediation.
• explainedHyde’s ciphered superscriptions — not a cipher at allresolved 15 Sept 20261659–60The four “undeciphered addresses” on Hyde’s letters to Barwick decode to nothing under the full Hyde–Barwick key printed in 1721, because, as the 1724 editor states, they were numbers “signifying nothing … only to puzzle the Enemy”.
• readThe Marqués del Carpio to Fuenmayor, 1677 — a Spanish ambassador’s syllabic cipher rebuilt and all ten letters readsolved 19 Sept 20261677Ten ciphered letters from Spain’s ambassador in Rome to its envoy in Denmark, eight deciphered in the margin without the key and two catalogued as non-decrypted. The key was rebuilt from the letters: numbers for letters and for syllables in rows of five, struck numbers for a second syllable table, and letter pairs that are syllables with a disguised consonant. It reads all ten: the French fleet at Messina and Catania, Orange raising the siege of Charleroi, Innocent XI refusing Carpio an audience, and Cardinal d’Estrées at Turin.
• readFeuquières to Catinat — the petit chiffre of the Pignerol governorssolved 16 Sept 20261691The 418-group despatch of 25 January 1691 that the 1819 editor of Catinat’s papers could not read and Bazeries read but never printed. The same edition prints a second letter in the same 367-group code, Louvois to d’Herleville of 6 September 1690, with its contemporary translation; the two letters share 72 groups, and hand alignment carried between them reads 586 of their 601 tokens. Feuquières’ plan for the surprise of Veillane: two roads into the town, eighty horse to Saint-Ambroise, an attack at two points, and the dragoons not to be let escape into the castle.
• readLouis XIV and Louvois to Catinat — seven Grand Chiffre despatches, and the decision to abandon Piedmontsolved 16 Sept 20261691The 1819 editor of Catinat’s papers printed seven court despatches of July–September 1691 in figures and could not read them; Bazeries rebuilt the code from them in 1893, printed two in clear for the Man in the Iron Mask, and stopped. His table reads all seven from the library’s OCR of the volume, 12,362 groups, checked against the page images; the two he printed agree at 98 and 99 %. The other five are read here for the first time, among them the King’s twenty-page letter of 14 September: bring the army back over the Alps, hold the passes, keep Carmagnole to cover the negotiation with the Pope, then burn it, and take Coni in the winter.
• readRobert Yard to the Earl of Manchester — two “undeciphered” letters of October 1699, read with the key from the same paperssolved 18 Sept 20261699Two letters from the Secretary of State’s office to the ambassador at Paris, 12 and 16 October 1699, catalogued at Yale as “almost entirely in cipher, undeciphered”; the Historical Manuscripts Commission said no key was known. The key was in the same Manchester papers: a printed code sheet with handwritten numbers, the = 454, as Tomokiyo had inferred. Checked on a sibling letter deciphered at the time, it reads both letters group for group, apart from five slips of Yard’s pen. Messengers at Dover wait for Mills and Lord Drummond’s priest, and Ireland and the north report a design carryed on very privately at Saint Germains.
• readCharles VI to Count Windischgrätz, 1720–1722 — the Emperor’s secret letters read with the two keys from the same archivesolved 19 Sept 20261720–22Four autograph letters of Charles VI to his envoy at The Hague, catalogued on DECODE as non-decrypted, with both keys filed beside them: the Chancellery’s 977-entry nomenclator and the Emperor’s private key of numbers and signs. Transcribed and applied, they read every cipher passage. After Alberoni’s fall the Emperor fears a French separate peace with Philip V through Morville, rejects Sardinia for Lorraine, claims the Montferrat equivalent and wants Portugal in the Alliance; in July 1720 Beretti Landi proposes a marriage between his children and Philip’s sons.
• already in printFrederick the Great to Louis Michell, 28 December 1751 — an intercepted Prussian cipher matched to the King’s printed minuteresolved 19 Sept 20261751A numerical nomenclator letter of Frederick II to his London resident, copied by the Dutch in The Hague and catalogued on DECODE as non-decrypted (R1957). Its clear opening, its length and the order of its repeated groups (1064 Vienne four times, 4310 Sardaigne three times, 3042 dessein) identify it as no. 5263 of the Politische Correspondenz, printed in 1882 from the minute: the King’s secret order to learn whether London knew of an Austro-Russian plan to crown Charles of Lorraine in Poland.
• solvedArmstrong to Madison — the coded postscriptsolved 14 Sept 20261808Forty-nine groups of the 1,600-element “THE = 972” diplomatic code, read in full after reconstructing 580 groups from the State Department’s own pencil decodes on NARA microfilm M34 roll 13.
• gaps readErving at Madrid (1807) and Adams at St Petersburg (1812) — two legation codespart read 18 Sept 20261807–12Two American diplomatic-code results from the State Department’s microfilmed despatch books. Erving’s private letter of 24 March 1807, printed from Madison’s own decode with the key “not found”: every hole closes from both received copies and Pinckney’s decoded despatches, and the holes turn out to be Madison’s slips. John Quincy Adams’s despatches are not in Armstrong’s code; his own (the = 1385) is rebuilt from the clerk’s decodes on NARA M35 reel 3, and the nine lines of No. 88 that Ford printed as not decyphered read, 124 of 134 groups: Count Rumyantsev’s strokes at Wilna.
• readCardinal Soglia to the nuncio Viale Prelà — a nomenclator with an alphabetical codesolved 18 Sept 20261848Rome, 15 June 1848: a ciphered dispatch to the nuncio at Innsbruck, intercepted at Turin and printed by Mazzini’s L’Italia del Popolo with a prize for its reader, posted on Cipherbrain in 2014 as unsolved. 5 is the word break, pairs of the other eight digits are letters and syllables, 8XXX is a one-part code in alphabetical order; a nomenclator annealer, checked first on a synthetic control, gave the first words. A counter-order: stay with the Emperor, do not ask for your passports. The nuncio had left Innsbruck the day before. Eleven code groups stay open.
• too short to breakThe Debosnys cryptograms — a French syllabary, and why it stops thereattempted 15 Sept 20261883A convicted wife-killer’s invented script, about 1,200 glyphs, unread since he was hanged. Three passages transcribed. The cipher poem is twenty lines of rhyming couplets, and line length and rhyme both make it a syllabary too short for any crib-free attack.
• solvedThe Swatow telegram to Sun Yat-sen — a systematic code condensersolved 15 Sept 20261916Twenty consonants, five vowels, ten-letter words: a code condenser over the standard Chinese telegraph code. The family of systematic tables is small enough to brute-force, and one key reads 41 characters: Mo Qingyu’s independence at Chaozhou and Sun’s men ordered out of the Swatow garrison headquarters, 3 April 1916.
• readSun Yat-sen’s circle on the wire, 1916–17 — the intercepts the Foreign Ministry could not readsolved 18 Sept 20261916–17The Ministry of Communications copied every telegram to or from Sun Yat-sen’s people for the Japanese Foreign Ministry, which filed them unread. Fifty-six condenser telegrams of March–May 1916 read under five more keys of the Swatow telegram’s family, each accepted only when it read telegrams it was not fitted to, with rhyme-day dates as an independent check. They cover the monarchy cancelled, money wired via Taiwan, the price of a Wuhan rising, Chen Qimei’s murder, and the Foreign Ministry’s own advice relayed back to Sun. Also a Hankow +111 telegram and Sun ↔ Dai Jitao, March 1917.
• scheme foundHuang Xing to Lin Hu and Li Genyuan — a kana condensersolved 15 Sept 20261916Listed as “solved but specific scheme unknown”: the Japanese Foreign Ministry filed a decode nobody could read, and an encipherment nobody could name. Both recovered, and then corrected from the frames: three kana carry one character deterministically from a private code, a superfluous kana had hidden the repeats, and the vowel is not free.
• 9 of 22 readThe ADFGVX residue of 1918 — twenty-two mutilated messages against known keyspart read 15 Sept 20261918Schmeh’s Top 50 lists twenty-two German radio messages of November 1918 as unsolved, but the keys were published in 2017 and the messages are garbled in transmission. The comment thread that read nine of them is tabulated for the first time, Lasry’s unpublished CHI key is rebuilt, his readers’ method re-derives the solved pages blind, and the ten never read resist every key and a key-free attack that fails its own control.
• no messageThe Chinese gold bar cryptograms — ten of everythingresolved 15 Sept 20261933Sixteen strings on seven bars said to certify $300,000,000, unread for ninety years. They contain almost exactly ten of every letter of the alphabet: chi-squared 1.25 against uniform where chance predicts 25. No cipher can flatten a distribution past what randomness allows, so there is nothing to read.
• no messageThe Roosevelt cryptogram — the numbers 1 to 52, written once eachresolved 16 Sept 20261935Three lines of digits above a threat to the President, reproduced by Friedman and unread since 1935. Read at glyph level they are a permutation of 1 to 52 with zero padding; the permutation drifts upward, runs 39 40 41 42 and falls into sixteen rising chains where a shuffle gives 26, the marks of a hand-written list. The ordered-key cipher readings that can be tested fail while matched controls are recovered.
• double pass not brokenGerman police double-box radiograms, 1942 — one key, not brokenattempted 18 Sept 20261942Six Ordnungspolizei radio messages from the Eastern Front in the Army’s double-box cipher, published by Frode Weierud and posted by Klaus Schmeh in 2020 as unsolved. One has a Bletchley decrypt: sixteen police killed on the Bobruisk–Mogilev road and the village of Borki razed. The five of 27 February 1942 turn out to share one key, 970 letters. Every ciphertext-only attack fails, on the real text and on a synthetic control of the same size, because the second pass through the same boxes leaves no gradient. A known-plaintext solver needs about 80 pairs, and the Borki decrypt does not align with its ciphertext.
• not a simple substitutionThe Copenhagen cryptogram — not a simple substitution of any language testedattempted 15 Sept 2026c.1950sThree lines found behind an 1835 portrait of a Danish general, 107 characters, unsolved at the American Cryptogram Association since the 1950s. Transcribed twice, attacked in ten languages under six reading conventions; the same solver recovers matched controls at 96 to 100 percent and the note never comes close.
• below unicity distanceThe Scorpion letters — two ciphers below the unicity distanceattempted 15 Sept 20261991Two Zodiac-style cryptograms sent to John Walsh in 1991, 70 and 180 symbols with 53 and 145 distinct. Both carry more key than the English text has redundancy, so fluent false solutions are guaranteed; matched controls produce them at 3 to 13 percent accuracy, and a claimed 2018 solution is tested against the same rule.
• the famous onesWhy the famous ciphers resistpart read 15 Sept 2026surveyKryptos, Voynich, Dorabella, Beale, Linear A, Phaistos, the pigeon message. Sorted by the actual reason each has held out: undeciphered writing systems, one that is information-theoretically secure, several too short for any answer to be provable, and at least two that were probably never enciphered.

02 Only in the notes

Results recorded in the repository README that have no page here yet. Each row links to the folder with the notes, transcriptions and code. The rows are read from the README at build time, so a result lands here as soon as it is logged there.

• attempted1520s superscript-digit ciphersnotes ↗1526–29Blocked: DECODE and BL images need login
• found solvedJean du Bellay → Montmorency, London, BnF fr. 3078 nos. 5, 17, 19 and fr. 3077 nos. 18, 23 (catalogue item 4)notes ↗1528–29All five in print in clear (17 Sept 2026): the notice hides the years; Le Grand's Histoire du divorce III (1688) prints the Court's readings of 15 June, 18 Sept, 17 and 27 Oct 1529, and fr. 3077 no. 23 is the 28 Oct 1528 letter whose fifteen cipher lines Bourrilly and de Vaissière read in 1905 with a key they recovered (Lasry 2022). The key re-derived here reads the 17 Oct letter (four pages of cipher, no interlinear) straight from the leaf, j'ay esté veoir le Cardinal en ses ennuis, où que l'y ay trouvé le plus grant exemple, as Le Grand prints it. By-product: fr. 3078 no. 3, 16 June 1529 (p. 21), is in no edition and DECODE calls it non-decrypted; its key was re-derived on the leaf from the decipherer's interlinear reading of the 22 June letter (B = o and the word signs bien, fault, paix added to Tomokiyo's table) and the block reads in stretches — semblant [que] madame ne fust … pour conclure … la [name] luy ayt par[lé] … [il] fault envoyé querir pouvoir, il ne fauldroyt oublyer a le faire — about 60 % of the signs. Scheurer 1969 unchecked
• attemptedCourt → Bernardin Bochetel, bishop of Rennes, BnF fr. 3181 f. 55 and 500 Colbert 390, 392 (catalogue item 11, class B)notes ↗1563Images fetched and Tomokiyo's 1561–64 key calibrated on the deciphered siblings; no continuous reading of the four target letters yet
• attemptedCatherine de Médicis → Philibert du Croc, Saint-Maur, 27 Apr 1567 (Destray 1924, facsimile plate)notes ↗1567Transcribed from the Gallica plate; a symbol cipher too short for an unconstrained solve unless the ticks turn out to be separators
• attemptedRegent Moray → John Wood, Add MS 32091 f. 213 (134 groups, 32 symbols)notes ↗1568Homophonic alphabet in Scots on the face of it, and a 5-gram Scots annealer reads 5 of 6 matched 134-letter controls, but the target gives no reading in Scots, English, French or Latin; no null, separator or crib is supported. The catalogue says the passage is Border news, i.e. names, and 13% of genuine Scots windows score below the solver's false optimum, so the result is undetermined rather than excluded. Needs the page (BL offline) or a second letter in the cipher
• attemptedLodovico Birago → Duke of Nevers, Saluzzo 13 Nov 1571 (BnF fr. 3251 f. 119)notes ↗1571Glyph-level transcription from Gallica: 483 digits, 16 marked digits, 9 inline wavy signs, 6 null letters. Variable-length prefix and suffix codes excluded against controls; the only consistent design is two-digit letters with one- or two-digit marked code groups (75 pairings, 228 tokens over 62 symbols), and at that homophony the annealer fails matched controls, so the target's negative is the method's limit; polyphonic single figures (Lucca-type) excluded with a method that reads its control 3/3; no sibling letter in the volume. Needs a sibling letter or a crib
• attemptedHenry III → Ségurnotes ↗1583–86Blocked on Gallica access; the sibling cipher's design is known
• attemptedSP 53/16 nos. 78-79 and SP 53/22 f. 52 (Mary Queen of Scots papers residue)notes ↗1585Symbol ciphers of the Paris-Rheims exile network in Tomokiyo's glyph-label transcription, 507 and 644 groups. Second session: the two letters share one key (10 of the 20 commonest symbols in common, 2.9 expected), so they pool to 1151 tokens; third session: a numba pipeline now reads clean 507- and 1151-token controls (and a 507-token control with word-signs), yet no. 78 shows no language basin in English, French, Latin, Italian or Spanish, no. 79 sits at random level, and pooling scores worse than either letter, so the same-key claim is withdrawn; a nulls-plus-nomenclator control is not read at this length, so that design stays open. f. 52 (84 tokens) is below unicity. Needs the images and the SP 53/22 keys
• partly readMayenne → Henri III and the Forget–Matignon letters, BnF fr. 15572 (catalogue item 12, class B)notes ↗1586f. 143 about half read (17 Sept 2026): 21 lines transcribed; the reading agrees with Tomokiyo's partial table and runs on past it (il se vouloit départir du roi de Navarre). The other nine leaves are transcription work.
• attemptedDoge Cicogna and Senate → Giovanni Mocenigo in France, subscribed Marco Ottobon, 27 Apr 1589 (BNE Mss/994 ff. 35-38; 1,528 tokens, 196 symbols)notes ↗1589Folios obtained from BNE Digital (130 dpi) and transcribed at first pass, about 70 % reliable per token. The letter is partly in clear: the Lyon postmaster forwarded only a few lines of the dispatch of 28 March, the rest and three other letters were intercepted, hence the cipher. Design: base letters a c d f g h with figures 1-99, letters, syllables and words mixed, the 1577-78 Zifra Prima type (not Franceschi's N.11, whose alphabet is confined to 1-20). Letter-only and letter-plus-wildcard annealers fail matched synthetic controls (0-9 % recovery even without transcription noise), so nothing is read. Route: the key sheets DECODE R1789/R1790, the register copy in ASVe Senato Secreta 1589, or a 300-dpi scan
• attemptedBlancmesnil → Nevers (fr. 3633 f. 24)notes ↗1589?No ciphertext online: fr. 3633 is not digitised (the source page's catalogue link goes to fr. 4736); the Potier-Nevers key of 1589 is the candidate once a copy exists; sibling letters with decipherments sit in fr. 3616/3621 (DECODE R9434-9448, images restricted)
• offline onlyMarco Ottobon → Giovanni Mocenigo, BNE Mss/994 ff. 34–38notes ↗1589One browser session: the manuscript is digitised (BNE Digital oid 0000174344; DECODE R2252, nine openings) but bne.es sits behind a Cloudflare Turnstile check that blocks scripts and automated browsers, and DECODE needs a login. About 6.5 pages of letter-plus-figure Venetian cipher, est. 2,500–4,000 tokens; the 1578–87 zifra granda syllabary is transcribed as a candidate key and the structural tests are written
• partly readBaudouin-Desportes → Aldobrandini and → Frachetta, Paris 22 July 1593, BnF fr. 3984 ff. 186, 189 (catalogue item 20, class A)notes ↗1593Not read (17 Sept 2026): the cipher is Tomokiyo's League polyphonic key (each glyph two letters), and the same-day letter to Lisieux (f. 188, deciphered on f. 184-185) gives 1,100 aligned glyphs of the hand; but a CNN trained on them reads only three glyphs in four on held-out lines, hand reading does no better, the polyphonic decoder needs over nine in ten, and the two letters share no verbatim passage with the Lisieux letter (alignment search against a shuffled control). c. 5,000 glyphs remain to be read; f. 177 (decipherment of f. 176) is the untapped training source
• attemptedCardinal de Joyeuse → Villars, Rome, 15 Feb 1594 (500 Colbert 33 f. 539)notes ↗1594Transcribed from Gallica: 358 tokens in three layers (81-sign homophonic alphabet, 48-sign marked syllabary, 29 code numbers); not Lasry's Senecey key from the same volume; the letter-layer anneal recovers 2-18 % of three matched controls and the target's seeds disagree
• attemptedCocquet → Mangot, Rome, Nov 1616 (Clairambault 369 f. 317)notes ↗1616Gate check: ~160 letter-shaped glyphs in nine runs inside clear French; none of the five period keys from the same volumes matches; du Croc regime
• offline onlyMaurice → Rupert; royalist intercepts, BL Add MS 72438notes ↗1645–46BL volumes digitised but offline since the 2023 cyber-attack
• attemptedCharles I, Isle of Wight lettersnotes ↗1648Two letters still unread; two candidate keys newly excluded
• attemptedThomas Urquhart's encrypted poemsnotes ↗17th c.Provenance objection to the Aug 2026 claim checked independently
• offline onlyCharles II → Duke of Hamiltonnotes ↗1650Copy order for NRS GD406/1/2197 (open)
• attemptedThurloe State Papers interceptsnotes ↗1653–56Four short pieces, all period keys fail
• attemptedLe Tellier → Castelnaunotes ↗1657Too short for an unconstrained syllabic solve
• attemptedColbert passages (Mélanges Colbert)notes ↗1665–75Read from the Gallica images: the Charost letter (Calais, 3 July 1675, not 1673) and the Gravel–Maulevrier copy (Mainz, 1674) are one key, Maulevrier's, 106 plain figures 52–489 between them. The printed catalogue located all 22 other Charost letters online and both Maulevrier letters: every one in clear. A monotone annealer reads matched controls of both strict alphabetical designs this office used and fails the target under each; looser designs fail their own controls. Probably a word nomenclator; needs the key or a clear copy of the news
• found solvedWilliam Perwich → Lord Arlington, Paris, 9 Apr 1670, TNA SP 78/129 f. 180notes ↗1670Broken twice in October 2025 (Matthew Brown; Lasry, Biermann and Tomokiyo): a columnar transposition with 20 columns and nulls. Reproduced here (solve20.py); this project's own earlier reading was wrong.
• attemptedLouis XIV → duc de Chaulnes, Rome (300-group one-part code)notes ↗1690Ciphertext verified from the page images and the code's ten-column Croissy design established, but 300 groups do not determine a 116-entry nomenclator: the annealer recovers 4-12% of a matched control and wrong keys score within noise of the true one. No printed plaintext found. Needs the minute in AE Rome Corr. 331-332
• offline onlyStepney → Manchester, Viennanotes ↗1702Stepney's office cipher in TNA SP 105/106 or BL Add MSS 7058–78
• offline onlyTorcy → Geertruidenberg plenipotentiaries; Villars → Polignac (BL Add MS 61575)notes ↗1710The ciphertext itself: Tomokiyo's transcription files are dead links, DECODE R8755/R8756 need a login, the BL volume is not digitised
• attemptedD'Estaing → Gérardnotes ↗1779217 tokens of a 600-code with no key material; needs the archive copy
• attemptedBerthier → Napoleon; letter to Marmontnotes ↗1807–12Neither attackable from the single printed source (Vilcoq 1969)
• attemptedArmstrong → Madison, 20 Feb 1808 (unique code, 369 groups to 1900, plus shorthand-like symbols)notes ↗1808Adjudication of the AFIO contest solution (May 2025): the 56-entry key covers 36 % of the groups, leaves 93 of its own 133 mapped occurrences unread, rests on five numbers absent from the letter, and fits its sentence worse than a key built by the same procedure on a shuffled ciphertext. Madison himself wrote in May 1808 that no such cipher was in the office, and the roll-13 pencil decodes are all THE=972, so they cannot supply it. The claim does not hold; the letter stays unsolved
• found solvedMilroy cipher telegrams (Cox 3 Dec 1862, Bascom 4 Dec 1861)notes ↗1861–62Solved by Richard Bean with Claude Opus 5 (2026): Stager cipher No. 7, six fixed columns, the indicator word giving the number of lines. Route spot-checked here from Milroy's own copy.
• attemptedCatokwacopa advertisementsnotes ↗1875Readings audited: which the letters force, which are guesses
• explainedBeale Paper no. 1notes ↗1885Fabrication; evidence in notes, book-cipher scan over Gutenberg negative
• in progress"BLUME SALAMANCA" telegrams, Zurich → Londonnotes ↗1937Transposition of telegraphic Spanish. Every single transposition and every double columnar with a second key of ten or fewer letters excluded against planted controls (lag scan plus exhaustive second-key enumeration). Open: double columnar with two long keys; needs the second telegram. Attempted, not solved
• attemptedZhongshan telegramsnotes ↗c.1938No corpus exists online; the premise of the list entry was wrong
• explainedD'Agapeyeff challenge ciphernotes ↗1939The ciphertext is not enciphered English
• attemptedWW2 censorship-manual steganogramsnotes ↗1940sBlocked on image resolution; TNA's digital copy is the same scan
• attemptedKoehler cryptograms (Abwehr)notes ↗1944Five short letter-cipher messages; skipped as intractable
• in progress"KR Blitz" Enigma message, OB Oberrhein → OKW, 10 Jan 1945 (TARGETS no. 19)notes ↗1945Ciphertext-only attack begun 16 Sept 2026; the first pass is partial and has produced no convincing plaintext
• found solvedFeynman ciphers nos. 2 and 3notes ↗1953Solved by David Vierra (May 2023): two substitution alphabets alternating word by word, even-length words reversed, boundaries dropped. Re-encryption verified here, 228 of 231 positions for no. 3.
• attemptedKaliningrad bottle postnotes ↗found 2015Blocker is transcription from two photographs, not cryptanalysis