At a glanceKey from adjacent plaintext · partial (60% of the cipher)
- The letters
- Peder Vibe (Copenhagen, later Bierreds and Flensborg) to Peder Juel, Danish resident in Stockholm, 1651 to 6 May 1655: eleven letters or pages in Danish with passages in a number cipher.
- Where
- Rigsarkivet, Copenhagen, TKUA, Speciel del, Sverige: Resident Peder Juels og Mogens Høegs gesandtskabsarkiver, Ordrer og breve 1648–55; Arkivalieronline unit 261583, images 77–93. Not on DECODE.
- The cipher
- Each cipher word begins with an opener (1–3 or 24–39) that sets a shift; the other numbers are letters of a 25-letter alphabet, shifted. Two- to four-digit code numbers stand between the words, by the sense for people and places.
- How the key was found
- On one letter, N. 319, someone at the time wrote letters over about a third of the cipher words. They gave the system and the opener table, which then decoded the other letters.
- The tests
- The first pre-registered test failed (margin 0.21, bar 0.30). A second, with the rule changed after that result, passed (margin 0.841). It validates the shift-per-word system, not the fixed table.
- What it says
- How to copy a seal from wax impressions; servants bribed for letters; fireships; an opened letter; something to lie in the Sound. Who does what is mostly behind the codes.
- Credit
- Transcription, key, tests and reading by Brice Dudout and his AI agents (pull request #410). The gloss on N. 319 is of the time; who wrote it is not known. The lead and the pointer to the glossed letter came from the project’s Rigsarkivet sweep.
- Still open
- All 73 code tokens (49 numbers); two cipher words; 14 words are only probable. No human has read the text.
01 The letters
Peder Vibe wrote to Peder Juel, the Danish resident in Stockholm, from Copenhagen and later from Bierreds and Flensborg. The letters lie in the resident’s embassy archive at the Rigsarkivet, numbered in an old hand. Most carry a few lines of numbers between the clear Danish.
In January 1649 Vibe had promised Juel a new alphabet (“dett ny Alphabet”) and gone on writing in the old cipher meanwhile. The letters on this page are the ones in that new alphabet: N. 316 (image 77, a slip) and N. 319 (images 78 and 79 left), both undated, which the contributor places in 1651; N. 334 (6 January 1653); N. 332, the end of a letter on image 83, N. 331, N. 333 and N. 335; N. 336 (Bierreds, 13 September 1654); N. 143 (1654); and N. 167 (Flensborg, 6 May 1655). The letters of 1649 to 1651 in the same file are in the old cipher, a different key, with its own page: Vibe to Juel, the old cipher.
One letter, N. 319, is the way in. Someone at the time wrote small letters over about a third of its cipher words:
02 The cipher
A cipher word is an opener followed by a body. The opener fixes a shift k, and each number of the body is the rank of a letter plus k.
- The alphabet has 25 letters: a to u without j and v, then w, x, y, z, ø (w = 21, x = 22, y = 23, z = 24, ø = 25). It does not wrap round.
- The opener table is the same in every letter: 24–29 give a shift of 0, 30–34 give 1, 35–39 give 2, and 1–3 give 3.
- In the body 32, 33 and 34 stand for 1, 2 and 3.
- A long word may be cut in pieces, each with its own opener.
The first word on the figure, 3.7.8.21.5.4.23, has the opener 3, so the shift is 3: 7 is d, 8 is e, 21 is s, 5 is b, 4 is a, 23 is u. It runs on in the next line and gives “desbauscheret”, as the gloss above it has it.
Between the cipher words stand two- to four-digit numbers, a nomenclator: 73 of them in these letters, 49 different numbers. 872 comes nine times; in the letter of 1655 a new list runs up to 1771. None has a value from a key. In January 1653 Vibe adds two numbers to the list in clear (349 and 561), but the names he gives them are hard to read and are left out of the reading. None of the 49 occurs in the old-cipher letters.
The key is key.tsv in the repository folder, released by the contributor under CC BY 4.0. A short script decodes the transcription with this one fixed table and no language model.
03 The tests: one failed, one passed
Brice Dudout’s agents wrote each test down and hashed it before opening the letters it would be run on, and ran each once. (The order of events is by the contributor’s record: the pull request is a single commit and the timestamps are his own.) The score is the mean log-probability per character of the decode under a 4-gram model of old Danish; the bar set in advance was the best control plus 0.30.
Test 1 failed. It declared that one opener table, that of N. 319, held for the whole file, with a 23-letter alphabet. The held-out set was three letters never opened: N. 330 (image 74), N. 316 (image 77) and N. 334 (image 79 right), 54 words and 335 numbers. The decode scored −3.068; the best of 600 controls scored −3.281. The margin was 0.21, under the bar of 0.30. The set was not measured again.
The rule was then changed, and the contributor says so. After test 1, N. 330 turned out to be in Vibe’s old cipher. Test 2 was written with a weaker claim: a shift per word, with the opener table not given but fitted on each letter. Every control got the same fitting.
Test 2 passed. The new held-out set was five images never opened (six letters or pieces), all in the new alphabet (images 81, 83, 84, 87 and 90; 55 words, 339 numbers). The decode scored −2.713; the best of 300 controls scored −3.555. The margin was 0.841. The raw decode also gave connected Danish sentences, the second criterion. One word of image 87 had been deciphered in the head while transcribing, before the hash; that is declared in the frozen file.
What test 2 shows is that the letters are written with a shift per word. It does not test the fixed table of section 02: the table was fitted afresh on each letter, and the 25-letter alphabet with w and ø was settled only after the measurement. The fixed table rests on other evidence:
- The gloss of the time on N. 319 agrees with the mechanical decode wherever it is legible.
- The tables fitted blind in test 2 agree with the table of N. 319 for most openers on images 83, 87 and 90, partly on image 84 and not on image 81, which has five cipher words; the contributor's summary is “four letters of five”.
- When the pull request was verified, both tests were run again and gave the same figures. The fixed table was then scored on the second held-out set with no fitting: −2.611 against a best control of −3.174, a margin of 0.563. Table and set had both been hashed before the set was opened.
- Also at verification, the three letters of test 1 were scored one by one. The two in the new alphabet scored above their controls (−2.353 against −2.961, and −2.632 against −3.094); the old-cipher letter scored −3.980 against −3.125 and sank the set. On the final transcription the fixed table scored −1.987 against a best of −2.568 for 200 random opener tables.
The checks made at verification came after the fact. They are controls, not pre-registered tests.
04 What the letters say
The cipher words give Danish that reads with the clear text. But the subject or object of most sentences is a code number, so the reading tells what was done and seldom who did it. In the quotations, capitals are cipher words, {braces} mark a probable word, and #876 is a code number.
N. 316, a slip. “TIL AT FAA #876 SIGNET RET UDSTUCKET er fornøden att #355 TRYCKES #876 ORDINARIE SIGNET UDI LACK to eller tre gange del KIENDELIG. Daa kand man DEREFTER LADE GIØRE IT ANDET som kand were IUST EFTER DET ANDET, att alt kand were WFORMERCHT.” To have the seal of #876 rightly engraved, #355 must press the ordinary seal of #876 in wax two or three times, clearly; then another can be made exactly like it, so that all stays unnoticed.
N. 319. Some time ago #545 debauched the boy of #818, who took a big packet of letters. #485 took his leave with the valet of #254, to do his best “til att DESBAUSCHERE MIN FORIGE {TIENER} CHARLES, OM HAND KUNDE FORSCHAFFE nogle aff #872 BREFUE”: to debauch my former servant Charles, to see if he could procure some letters of #872. “Quo poterant ibant, quo non poterant ibi stabant.” Work is going on to intercept letters of #872 of importance.
N. 334, 6 January 1653. #353 leaves with the first occasion for #770, and #397 likewise. Vibe will say nothing of their instruction; something can be dealt with at a border meeting (“GRENTZE MØDE”).
Image 83, the end of a letter. Würtz (with the code #623 before his name) will persuade #454 that #770 has resolved something against #217. “Der sigis dog for wiss att HAND SKAL WERE NYLIGEN NOBILITERET aff #222 UDI #770 OCH FAAET GODS”: it is said for certain that he has lately been ennobled by #222 in #770 and received estates, so that one does not know how far to trust him.
N. 333. Since #222 is having so many fireships built (“BRANDT #791”), it would be needful to learn by what name or sign each can be told from the other ships.
N. 332, N. 331 and N. 335 are short or broken: a young lady Hedwig and courtiers (N. 332); Würtz’s credit in Sweden (N. 331); letters, copies and a report of #872 (N. 335), of which only single words are in cipher.
N. 336, 13 September 1654. Vibe asks what is decided according to the order #397 has received, and “HUAD #879 BREF SOM BLEF OPBRUT INDHOLT”, what the letter of #879 that was opened contained.
N. 143, 1654. He begs Juel, by their friendship, to tell him “HUO DEN ER SOM #879 TROER BEST”, who it is that #879 trusts most, and swears in French to keep it secret.
N. 167, 6 May 1655. A passage almost wholly in code numbers, 24 of them: #1590 has a design, which leaves the States General “NOGET PERPLEX”, somewhat perplexed, and something is to “LIGGE … UDI SUNDET”, lie in the Sound.
The contributor notes two guesses that the sense invites, 770 for Sweden and 222 for Queen Christina. No key supports them and they are not part of the reading.
05 What remains open
The count uses one rule. A cipher word counts when every number is legible and the word is Danish (certain, 132) or when one number had to be corrected for a stated reason (probable, 14). A word supplied from the sense (1) or unresolved (2) is a hole, and so is every code number (73).
- Cipher words given a value: 146 of 149, 98.0%.
- Read share with every code as a hole: 132 of 222, 59.5% (certain words only; probable, restored and unresolved words and every code are holes). With the 14 probable words it is 146 of 222, 65.8%. That is under the 95% bar, so the text is deciphered in part. The pull request gave one figure for both of Vibe’s ciphers together, 71.8%.
- Two pieces are under 95% on cipher words alone: the end of a letter on image 83 (17 of 18) and N. 167 (12 of 13).
Open pieces:
- The 73 code tokens, 49 numbers. No code list is in the file, and the gloss on N. 319 covers letters only. Juel’s replies, the rest of NKS 1299 e at the Royal Library, or a code list in the Juel papers could give them.
- Two unresolved cipher words: on image 78,
31.34.26.13.20(“bømt”, read clearly, glossed only “°m t” at the time); on image 93,30.?.9.2.14=5.6.20(a blot after the opener). - One restored word: on image 83 the cipher reads “hide” where the sense wants “hende”, her.
- The 14 probable words. Each rests on a corrected number, a re-divided tick, a supplied letter or an unusual spelling, listed with its reason in the reading file. None was compared with the image by a second reader.
- Prior art. The catalogue vetting found nothing on DECODE, at Cryptiana or in George Lasry’s lists, and the contributor found no edition of the cipher passages. Neither search was repeated when the pull request was verified, and the Royal Library and the Danish printed collections were not searched at length.
- All readers so far were AI agents working from the microfilm scans online. At verification about 95 numbers of image 83 were compared with the image and agreed; no person has checked a passage.
Three letters of 1648 from a writer named Krag, in the same file, are in a third system that is not solved. They are kept as a lead, not on this page.
Full texts
The transcription and the decryption of every document on this page, so that the reading can be used and checked. Short files are shown here; all are in the repository.
Peder Vibe to Peder Juel, 1651-1655, cipher passages in the new alphabet of 11 letters or pages read in part
Secret passages of Vibe's letters from Copenhagen, Bierreds and Flensborg, 1651 to May 1655: a seal of #876 to be copied from wax impressions; a servant debauched to get a packet of letters, and work to intercept letters of #872; #353 and #397 leaving for #770 with an instruction, and a border meeting (1653); Würtz said to be ennobled by #222; fireships built by #222; an opened letter of #879 (1654); a design that leaves the States General perplexed, and something to lie in the Sound (1655). The 49 code numbers that name the actors are not identified.
Builds coherent sentences, but under the read bar of 95% of words read as sense (60% here); the holes are marked in the reading.
Transcription
# vibejuel1651 - transcription of the cipher passages of the letters of Peder Vibe to Peder Juel (Danish resident in Stockholm)
# in the NEW alphabet ("ny Alphabet"), 1651 to 6 May 1655 (Rigsarkivet, TKUA, Speciel del, Sverige, Resident Peder Juels og
# Mogens Hoegs gesandtskabsarkiver, Ordrer og breve 1648-55; Arkivalieronline unit 261583, images 77-93). Part of catalogue 2527.
# Conventions: the numbers of one cipher word are joined by "." (the scribe marks them off with small ticks below the line);
# "=" at the end of a token = the word goes on after a line break or with a new opener; C<n> = nomenclator code (3 or 4 digits,
# sometimes 2); [ ... ] = clear text of the letter between the cipher words, abridged where marked "..." (notes in it are
# in English: struck out, blot, tear, hole in the paper); "!" = doubtful word.
# "#A" and "#K" lines give the alphabet and opener table used by scripts/decode.py (see key.tsv). "#S" starts a letter or page.
# Transcribed from the Arkivalieronline images by the contributor's agents (Brice Dudout, 11 Nov 2026), number by number; re-read
# at native zoom after the measurements. The version-0 transcriptions of the two held-out sets, frozen before any decryption, are in
# targets/vibejuel1651/frozen/working/transcr/ with their hashes in targets/vibejuel1651/frozen/protocole/.
# Split by key on 12 Nov 2026 (one key, one entry): the letters in the old cipher (1649-1651) are in targets/vibejuel1649/; the three Krag
# letters of 1648 (another system, not solved) are in research/sweep_2026-11-11_tkua_paging/krag1648_transcription.txt.
# Integration fix (12 Nov 2026): the code 545 that opens the passage of N. 319 (image 78) is now a C token (it stood inside the clear bracket).
# Credit: Rigsarkivet, Arkivalieronline.
#S image 77: N. 316, slip (held-out set of test 1) - ny Alphabet
#A B
#K {24:0,25:0,26:0,27:0,28:0,29:0,30:1,31:1,32:1,33:1,34:1,35:2,36:2,37:2,38:2,39:2,1:3,2:3,3:3}
3.22.12.14 2.4.22 30.7.33.33 C876 35.20.11.9=
15.7.21 37.19.7.21 32.21.5.19.20.21.4.11.6.20 [er]
[fornøden att] C355 31.20.18.24.4.11.6.19 C876
36.16.19.6.11.15.34.19.11.7 1.21.12.10.16.8.22 30.21.5.10
37.13.34.5.12 [to eller tre gange del]
3.13.12.8.16.7.8.14.12.10 [Daa kand man]
31.5.6.18.6.7.20.6.18 35.13.34.6.7 1.10.12.28.20.8
36.11.21 30.33.14.5.6.20 [som kand were]
37.11.22.20.21 2.8.9.22.8.20 32.5.6.20
37.34.15.6.7.21 [att alt kand were]
35.23.8.16.19.14.7.19.5.10.21
#S image 78: N. 319 (partial interlinear decipherment of the time; development set) - ny Alphabet
#A B
#K {24:0,25:0,26:0,27:0,28:0,29:0,30:1,31:1,32:1,33:1,34:1,35:2,36:2,37:2,38:2,39:2,1:3,2:3,3:3}
C545 [haffuer for nogen tid siden] 3.7.8.21.5.4.23=
21.6.11.8.20.8.22 C818 30.5.18.6.14.8 [som tog itt]
35.20.21.16.19.21 37.19.34.5.18.22.7.21 32.34.18.6.7=
21.6 [der] C818 [dett formaen vriste hand]
31.9.10.6.13.13.6.12.19.8.6.14 2.5.14.20.22 [om skiønt]
3.7.8.20 31.7.33.14.20.10.19 30.10.14.20.6.20 [struck out : 366…]
36.6.7.19 35.11.4.13.34.15.21 [som hannem]
32.11.21.14.5.6 31.21.6.18.6 30.16.18.6.10.21.5.10=
4.6.18.12.10.4 [Maaske aff] C873 [og aff] C479
3.5.20.8.9.23.8 [war der nogle mögel] 2.24=
9.17.20.21.6.11.4.15.8.22.8 C485 [tog den abschied]
[med] C254 [struck out] 1.6.4.15.15.8.20.22.12.8.16.8.20 [att]
[giöre sitt beste til att] 30.5.6.19.34.33.21=
19.4.9.6.18.6 31.13.10.14 3.9.17.20.12.10.8 35.21.11.17=
15.7.19 37.5.10.34.19.13.7.20 2.17.15 32.9.33.14.5
30.11.21.14.5.6 31.7.15.18.19.4.9.33.7.7.6 [nogle aff]
C872 31.34.18.6.7.21.6 [Quo poterant ibant quo non poterant ibi stabant. Man arbeder nu paa att] 37.7.19.21.34.17.7 35.15.16=
9.13.7 [aff] C872 36.4.19.7.8.22.7 3.4.9 2.12.15=
18.17.20.22.4.6.8 [huor til Brud eller giffuer]
[lejlighed] C872 30.20.33.8 [bliffuer wel ihr]
31.34.26.13.20 32.9.6.18.18.6.5.33.8 [Mens der=]
31.5.6.14.14.6
#S image 79 left: end of N. 319 - ny Alphabet
#A B
#K {24:0,25:0,26:0,27:0,28:0,29:0,30:1,31:1,32:1,33:1,34:1,35:2,36:2,37:2,38:2,39:2,1:3,2:3,3:3}
[som man kand faa nogen] 36.34.34.19.20.34.9
37.21.11.13 35.34.21 2.22.4.10.8 3.11.4.16.7.21
1.10.17.7.21 30.10 [struck out : 3?3] C73 [da schreff dett wiss]
#S image 79 right: N. 334 ('N° 52'), ends on image 80, Copenhagen 6 Jan 1653 (held-out set of test 1) - ny Alphabet
#A B
#K {24:0,25:0,26:0,27:0,28:0,29:0,30:1,31:1,32:1,33:1,34:1,35:2,36:2,37:2,38:2,39:2,1:3,2:3,3:3}
[Kiere Peder Jüll ... Da drager] C353 [med dett första]
3.22.12.14 C770 C397 [er wdi lige maade]
C67 30.13.6.5 C353 [Jeg will her intet formelde om] 35.6.7.19.11.20 2.12.16=
21.22.20.23.6.22.12.17.16 [effter som dett dig bliffuer nochsom Communiceret ... begynde] C462 [hvilche uden wed]
31.8.18.6.14.20.25.6 36.14.27.6.7 [kand afhandlis ... 353 Refererer wel 397 alle di Consilia som forfattis wdi di sager]
['fun' struck out] [mi] 2.11.8.20 37.15.7.9.16.5.11.7.19.11.20 [huorfor altingest remitteris der hend. 93 bliffuer wed hands forige ampt men 94 bliffuer employeret wdi en honorable]
#S image 81: N. 332 ('N° 56') (held-out set of test 2) - ny Alphabet
#A B
#K {24:0,25:0,26:0,27:0,28:0,29:0,30:1,31:1,32:1,33:1,34:1,35:2,36:2,37:2,38:2,39:2,1:3,2:3,3:3}
C876 [och] 3.9.20.28.13.8.16 30.9.6.5.22.10.8
[om di haffuer och weret] 37.11.4.13.34.15.21 31.5.10 2.6.17.23.20=
22.12.21.4.16.8.20 [som haffuer giort …]
#S image 83: left, end of a letter; right, N. 331 ('N° 60') (held-out set of test 2) - ny Alphabet
#A B
#K {24:0,25:0,26:0,27:0,28:0,29:0,30:1,31:1,32:1,33:1,34:1,35:2,36:2,37:2,38:2,39:2,1:3,2:3,3:3}
C623 3.24.23.20.22.27 [aff] C376 C357 [tienner och Confident] 30.22.10.12 35.17.7.19.20.22.34=
6.7.19.7 C454 [att] C770 37.10.34.8.22.7.19 2.20.8.21.17.14.23.8.20.8.22 C463 31.13.15.5 C217 [Der sigis dog for wiss att]
1.11.4.16.7 3.21.13.4.14 2.24.8.20.8 30.14.24.12.10.8.6.14
32.14.15.34.10.12.10.20.6.18.6.20 [aff] C222 37.22.6.11 C770 3.17.6.11 37.8.34.34=
7.21 30.8.15.5.19 [aff] !1.11.12.7.8 [huoraff=ere man icke wed huorledis man skall her] 37.21.19.16.7 35.10.34.15.15.7.14
[huilchet awiserer mig om] C620 30.22.21.18.20.25 [Credit och æstimation wdi Swerrig]
[Interim wil ieg ickun lade dig wide att] C872 3.18.20.17.6.23=
20.4.22.17.20 C95 [en forschrifft til] C583
#S image 84 right: N. 333 (held-out set of test 2) - ny Alphabet
#A B
#K {24:0,25:0,26:0,27:0,28:0,29:0,30:1,31:1,32:1,33:1,34:1,35:2,36:2,37:2,38:2,39:2,1:3,2:3,3:3}
[Och efftersom] C222 [lader giöre saa mange]
3.5.20.4.16.7.22 C791 [da war dett wel fornöden att kunde erfahre och wide huad nassuen eller tegnn en huer aff disse]
35.4.19.34.15.6.21 C791 [haffuer derhued att erkiende dennem fra di andre] 30.19.11.10=
34.6 [thi der er ingen tuiffuel paa att di bliffuer wel saaledis beschaffenatt som di andre] 2.21.13.12.5.8 [och maa der for tegnis ...]
[... wiss Resolution om] C343 [will efterkomme]
#S image 87 right: N. 335 ('Her besante') (held-out set of test 2) - ny Alphabet
#A B
#K {24:0,25:0,26:0,27:0,28:0,29:0,30:1,31:1,32:1,33:1,34:1,35:2,36:2,37:2,38:2,39:2,1:3,2:3,3:3}
[Jeg will icke underlade att lade dig wide att] 2.5.20.8 32.7.22.6
[ny] 35.17.16.20.21 [ere och til]
30.9.6.14.5.6 [komme fra] C872
31.7.10.14.5.6.19 [huad passerer …]
3.8.7 30.6.18.19 [och] C872 [relation findis …]
37.20.12.19.11.8.23.7.19 [att] C601 C397 [skal haffue …] C872
36.5.16.17.11.7.19 [aff …]
3.5.20.8.24.8 C872 [… att] C397 [… att ingen]
1.4.6.22.8 30.19.11.18.10.7.20= 12.10.8 [bleff giffuen]
#S image 90 left: N. 336, Bierreds 13 Sep 1654 (held-out set of test 2) - ny Alphabet
#A B
#K {24:0,25:0,26:0,27:0,28:0,29:0,30:1,31:1,32:1,33:1,34:1,35:2,36:2,37:2,38:2,39:2,1:3,2:3,3:3}
[… Lad mig wide huad] 37.14.8.6 35.10.34.15
3.16.8.15 2.21.14.23.22.12.21 [effter den] 30.34.6=
7.33.12.21.14.8 C397 36.6.7.19.16.14 1.5.8.13.17.15
30.13.6.20 [haffuer. Dißligest] 35.10.22.34.6
C879 37.4.19.7.8 31.19.15.13 3.5.14.8.9 30.15.16=
34.18.21.20 32.10.14.5.9.15.12.20 [Jeg wed icke …]
#S image 91 right: N. 143 / 321, 1654 - ny Alphabet
#A B
#K {24:0,25:0,26:0,27:0,28:0,29:0,30:1,31:1,32:1,33:1,34:1,35:2,36:2,37:2,38:2,39:2,1:3,2:3,3:3}
[… wed woris fortroende wenskab att du lader mig wide] 3.11.23.17
2.7.8.16 30.6.18 32.19.15.13 C879 35.21.19.16=
7.19 37.4.7.20.21 1.8.20 31.22.6.18.19.20
36.16.5.10 C872 37.4.7.20.21.7 3.24.8.16
#S image 93 right: N. 167 / 337, Flensborg 6 May 1655 (new nomenclator of 3-4 digits) - ny Alphabet
#A B
#K {24:0,25:0,26:0,27:0,28:0,29:0,30:1,31:1,32:1,33:1,34:1,35:2,36:2,37:2,38:2,39:2,1:3,2:3,3:3}
[… forsichert(?) wdi werden] C349 C763 [aviseris wi for wist att] C1590 [haffuer anslag] C1274 !30.9.2.14=
5.6.20 C1195 C671 [Huilchet giør Gen. Stat.] 35.15.16=
9.7.21 37.17.7.19.17.13.7.24 2.9.17.20 C430 C889 C428
C487 C770 [och begynder att] C467 31.9.21.15.18=
12.6.5.10.19 [dette] C1588 [deffsing(?)] C978 32.7.15.18=
9.10.14.5.18.10.19 C797 C467 [man, om man] C798
2.21.13.4.14 35.17.22.20.20.7.19.7 36.6.7.14 [att]
C428 C126 C1135 C1590 C1771 1.24.4.16.7.21 [och]
3.14.12.10.10.8 C448 C600 35.22.6.11 36.20.22.15=
6.7.21 [Huorom ieg begierer din betenckning …]Decryption
# vibejuel1651 - raw decryption of transcription.txt, letter by letter (Vibe to Juel, new alphabet, 1651-1655)
# Deterministic decode by scripts/decode.py with the fixed table of key.tsv (no language model, no correction):
# python3 scripts/decode.py transcription.txt
# CAPITALS = decoded cipher words; {n} = nomenclator code (not identified); [ ... ] = clear text of the letter; ?n = number out of the
# alphabet. Misread numbers are left as they come out (e.g. RACQUET for pacquet, BLRT for bort on image 78); the reading by sense, with every correction justified, is reading.md.
== image 77: N. 316, slip (held-out set of test 1) - ny Alphabet
TIL AT FAA {876}
SIGNET RET UDSTUCKET [er]
[fornøden att] {355} TRYCKES {876}
ORDINARIE SIGNET UDI
LACK [to eller tre gange del]
KIENDELIG [Daa kand man]
DEREFTER LADE GIØRE
IT ANDET [som kand were]
IUST EFTER DET
ANDET [att alt kand were]
WFORMERCHT
== image 78: N. 319 (partial interlinear decipherment of the time; development set) - ny Alphabet
{545} [haffuer for nogen tid siden]
DESBAUSCHERET {818} DRENG [som tog itt]
STORT RACQUET
BREFUE [der] {818} [dett formaen vriste hand]
HIEMMELSGEN BLRT [om skiønt]
DER FANTIS INTET [struck out : 366…]
DER IBLANT [som hannem]
KUNDE UERE
PREIUDICERLIC [Maaske aff] {873} [og aff] {479}
BREFUE [war der nogle mögel]
WFORSCHAMETE {485} [tog den abschied]
[med] {254} [struck out] CAMMERTIENER [att]
[giöre sitt beste til att]
DESBAUSCHERE MIN FORIGE
TIPNER CHARLES OM HAND
KUNDE FORSCHAFFE [nogle aff]
{872} BREFUE [Quo poterant ibant quo non poterant ibi stabant. Man arbeder nu paa att] ERTAPE
NOGLE [aff] {872} BREFUE AF
IMPORTACE [huor til Brud eller giffuer]
[lejlighed] {872} TAG [bliffuer wel ihr]
BØMT HERREDAG [Mens der=]
DENNE
== image 79 left: end of N. 319 - ny Alphabet
[som man kand faa nogen] AARSAG
TIL AT TAGE HANDS
GODS I [struck out : 3?3] {73} [da schreff dett wiss]
== image 79 right: N. 334 ('N° 52'), ends on image 80, Copenhagen 6 Jan 1653 (held-out set of test 1) - ny Alphabet
[Kiere Peder Jüll ... Da drager] {353} [med dett första]
TIL {770} {397} [er wdi lige maade]
{67} MED {353} [Jeg will her intet formelde om] DERIS
INSTRUCTION [effter som dett dig bliffuer nochsom Communiceret ... begynde] {462} [hvilche uden wed]
GRENTZE MØDE [kand afhandlis ... 353 Refererer wel 397 alle di Consilia som forfattis wdi di sager]
['fun' struck out] [mi] HER NEGOCIERIS [huorfor altingest remitteris der hend. 93 bliffuer wed hands forige ampt men 94 bliffuer employeret wdi en honorable]
== image 81: N. 332 ('N° 56') (held-out set of test 2) - ny Alphabet
{876} [och] FRØKEN HEDWIG
[om di haffuer och weret] IBLANT DI
COURTISANER [som haffuer giort …]
== image 83: left, end of a letter; right, N. 331 ('N° 60') (held-out set of test 2) - ny Alphabet
{623} WURTZ [aff] {376} {357} [tienner och Confident] WIL
PERSUADERE {454} [att] {770} HAFUER RESOLUERET {463} MOD {217} [Der sigis dog for wiss att]
HAND SKAL WERE NYLIGEN
NOBILITERET [aff] {222} UDI {770} OCH
FAAET GODS [aff] HIDE [huoraff=ere man icke wed huorledis man skall her] TROE HANNEM
[huilchet awiserer mig om] {620} WURTZ [Credit och æstimation wdi Swerrig]
[Interim wil ieg ickun lade dig wide att] {872}
PROCURATOR {95} [en forschrifft til] {583}
== image 84 right: N. 333 (held-out set of test 2) - ny Alphabet
[Och efftersom] {222} [lader giöre saa mange]
BRANDT {791} [da war dett wel fornöden att kunde erfahre och wide huad nassuen eller tegnn en huer aff disse]
BRANDT {791} [haffuer derhued att erkiende dennem fra di andre]
SKIBE [thi der er ingen tuiffuel paa att di bliffuer wel saaledis beschaffenatt som di andre] SKIBE [och maa der for tegnis ...]
[... wiss Resolution om] {343} [will efterkomme]
== image 87 right: N. 335 ('Her besante') (held-out set of test 2) - ny Alphabet
[Jeg will icke underlade att lade dig wide att] BRE FWE
[ny] POST [ere och til]
HENDE [komme fra] {872}
FINDES [huad passerer …]
ED ERS [och] {872} [relation findis …]
SKRIFWER [att] {601} {397} [skal haffue …] {872}
COPIER [aff …]
BREWE {872} [… att] {397} [… att ingen]
ACTE SKRIFTLIG [bleff giffuen]
== image 90 left: N. 336, Bierreds 13 Sep 1654 (held-out set of test 2) - ny Alphabet
[… Lad mig wide huad] MFD HAN
NEM SLUTIS [effter den]
BEFALUNG {397} DEROM BEKOM
MET [haffuer. Dißligest] HUAD
{879} BREF SOM BLEF
OPBRUT INDHOLT [Jeg wed icke …]
== image 91 right: N. 143 / 321, 1654 - ny Alphabet
[… wed woris fortroende wenskab att du lader mig wide] HUO
DEN ER SOM {879}
TROER BEST ER WERST
OCH {872} BESTE WEN
== image 93 right: N. 167 / 337, Flensborg 6 May 1655 (new nomenclator of 3-4 digits) - ny Alphabet
[… forsichert(?) wdi werden] {349} {763} [aviseris wi for wist att] {1590} [haffuer anslag] {1274}
HANDET {1195} {671} [Huilchet giør Gen. Stat.]
NOGET PERPLEX FOR {430} {889} {428}
{487} {770} [och begynder att] {467}
HUORLEDIS [dette] {1588} [deffsing(?)] {978}
FORHINDRIS {797} {467} [man, om man] {798}
SKAL PUSSERE DEM [att]
{428} {126} {1135} {1590} {1771} WANDS [och]
LIGGE {448} {600} UDI
SUNDET [Huorom ieg begierer din betenckning …]Reading
# vibejuel1651 - reading of the cipher passages in the new alphabet (Danish, spelling of the letters)
Letters of Peder Vibe to Peder Juel, 1651 to 6 May 1655, in the new alphabet ("ny Alphabet"; Arkivalieronline unit 261583, images
77-93). Read by sense from `decryption.txt` with `key.tsv` (Brice Dudout's agents, PR #410), after the measurements
(`validation.txt`). The re-reading of the held-out sets was done AFTER the measurements; the measurements were not run again. The
letters of 1649-1651 in the old cipher are in `targets/vibejuel1649/reading.md`.
Marks: CAPITALS = cipher words, certain (every number read, the word is Danish of the time); {X} = probable (one number corrected
with a reason taken from the image, or an old spelling; see the list); [x] = restored (word supplied from the sense, not counted);
<...> = unresolved; #211 = nomenclator code, not identified; lower case = clear text of the letter, abridged with "...". Word
division and punctuation are the contributor's. Spelling of the time kept (u for v, w for v, ff, hafuer).
## Text
**Image 77, N. 316 (slip).** TIL AT FAA #876 SIGNET RET UDSTUCKET er fornøden att #355 TRYCKES #876 ORDINARIE SIGNET UDI LACK to
eller tre gange del KIENDELIG. Daa kand man DEREFTER LADE GIØRE IT ANDET som kand were IUST EFTER DET ANDET, att alt kand were
WFORMERCHT.
**Images 78-79 left, N. 319 (partial interlinear decipherment of the time).** #545 haffuer for nogen tid siden DESBAUSCHERET #818
DRENG, som tog itt STORT {PACQUET} BREFUE; der #818 dett formaen vriste hand {HIEMMELIGEN} {BORT}, om skiønt DER FANTIS INTET DER
IBLANT som hannem KUNDE UERE {PREIUDICERLIG}. Maaske aff #873 og aff #479 BREFUE war der nogle mögel. WFORSCHAMETE #485 tog den
abschied med #254 CAMMERTIENER att giöre sitt beste til att DESBAUSCHERE MIN FORIGE {TIENER} CHARLES, OM HAND KUNDE FORSCHAFFE
nogle aff #872 BREFUE. Quo poterant ibant, quo non poterant ibi stabant. Man arbeder nu paa att ERTAPE NOGLE aff #872 BREFUE AF
{IMPORTANCE}, huor til Brud eller giffuer lejlighed #872 TAG bliffuer wel ihr <31.34.26.13.20> HERREDAG. Mens der DENNE ... som man
kand faa nogen AARSAG TIL AT TAGE HANDS GODS I #73, da schreff dett wiss.
**Image 79 right, N. 334 ("N° 52"), Copenhagen 6 Jan 1653 (date on image 80).** ... Da drager #353 med dett första TIL #770; #397
er wdi lige maade #67 MED #353. Jeg will her intet formelde om DERIS {INSTRUCTION}, effter som dett dig bliffuer nochsom
communiceret ... begynde #462, hvilche uden wed GRENTZE MØDE kand afhandlis ... HER NEGOCIERIS, huorfor altingest remitteris der
hend.
**Image 81, N. 332 ("N° 56").** #876 och FRØKEN HEDWIG, om di haffuer och weret IBLANT DI {COURTISANER} som haffuer giort ...
**Image 83 left, end of a letter.** #623 WURTZ aff #376 #357 tienner och Confident WIL PERSUADERE #454 att #770 HAFUER RESOLUERET
#463 MOD #217. Der sigis dog for wiss att HAND SKAL WERE NYLIGEN NOBILITERET aff #222 UDI #770 OCH FAAET GODS aff [hende], huoraff
ere man icke wed huorledis man skall her TROE HANNEM.
**Image 83 right, N. 331 ("N° 60").** ... huilchet awiserer mig om #620 WURTZ Credit och æstimation wdi Swerrig. Interim wil ieg
ickun lade dig wide att #872 {PROCURATOR} #95 en forschrifft til #583.
**Image 84 right, N. 333.** Och efftersom #222 lader giöre saa mange BRANDT #791, da war dett wel fornöden att kunde erfahre och
wide huad nassuen eller tegnn en huer aff disse BRANDT #791 haffuer derhued att erkiende dennem fra di andre SKIBE, thi der er
ingen tuiffuel paa att di bliffuer wel saaledis beschaffenatt som di andre SKIBE ...
**Image 87 right, N. 335 ("Her besante").** Jeg will icke underlade att lade dig wide att {BREFWE} ny POST ere och til HENDE komme
fra #872 ... FINDES huad passerer ... EDERS och #872 relation findis ... SKRIFWER att #601 #397 skal haffue ... #872 COPIER aff ...
BREWE #872 ... att #397 ... att ingen ACTE SKRIFTLIG bleff giffuen.
**Image 90 left, N. 336, Bierreds 13 Sep 1654.** ... Lad mig wide huad {MED} HANNEM SLUTIS effter den BEFALUNG #397 DEROM
BEKOMMET haffuer. Dißligest HUAD #879 BREF SOM BLEF OPBRUT INDHOLT. Jeg wed icke ...
**Image 91 right, N. 143 / 321, 1654.** ... Jeg beder och formaner dig ... wed woris fortroende wenskab att du lader mig wide HUO
DEN ER SOM #879 TROER BEST, ER WERST OCH #872 BESTE {WEN}. Je vous jure par le salut de mon ame, de tenir cela secret, och begierer
ieg dette att wide allenist til att menagere herefter mit wenskab.
**Image 93 right, N. 167 / 337, Flensborg 6 May 1655 (new nomenclator).** ... #349 #763 aviseris wi for wist att #1590 haffuer
anslag #1274 <30.?.9.2.14=5.6.20> #1195 #671. Huilchet giør Gen. Stat. NOGET PERPLEX FOR #430 #889 #428 #487 #770, och begynder att
#467 HUORLEDIS dette #1588 deffsing (?) #978 FORHINDRIS. #797 #467 man, om man #798 {SKAL} {PASSERE} DEM att #428 #126 #1135
#1590 #1771 WANDS och LIGGE #448 #600 UDI SUNDET. Huorom ieg begierer din betenckning ...
## Words
Unit: one cipher word spelled in letters (split words joined). Read = certain + probable; restored and unresolved words are holes;
every nomenclator code is a hole too, counted apart in the last column and added in the second figure below. One counting rule
for the whole target.
| Image | Letter | Words | Certain | Probable | Restored | Unresolved | C+P | Codes |
|---|---|---|---|---|---|---|---|---|
| 77 | N. 316 (held out, test 1) | 22 | 22 | 0 | 0 | 0 | 100.0% | 3 |
| 78-79 l. | N. 319 (development, glossed) | 44 | 37 | 6 | 0 | 1 | 97.7% | 11 |
| 79 r. | N. 334, 6 Jan 1653 (held out, test 1) | 8 | 7 | 1 | 0 | 0 | 100.0% | 6 |
| 81 | N. 332 (held out, test 2) | 5 | 4 | 1 | 0 | 0 | 100.0% | 1 |
| 83 l. | end of a letter (held out, test 2) | 18 | 17 | 0 | 1 | 0 | 94.4% | 9 |
| 83 r. | N. 331 (held out, test 2) | 2 | 1 | 1 | 0 | 0 | 100.0% | 4 |
| 84 r. | N. 333 (held out, test 2) | 4 | 4 | 0 | 0 | 0 | 100.0% | 4 |
| 87 r. | N. 335 (held out, test 2) | 10 | 9 | 1 | 0 | 0 | 100.0% | 7 |
| 90 l. | N. 336, 13 Sep 1654 (held out, test 2) | 12 | 11 | 1 | 0 | 0 | 100.0% | 2 |
| 91 r. | N. 143 / 321, 1654 | 11 | 10 | 1 | 0 | 0 | 100.0% | 2 |
| 93 r. | N. 167 / 337, 6 May 1655 | 13 | 10 | 2 | 0 | 1 | 92.3% | 24 |
| **Total** | 11 letters or pages | **149** | **132** | **14** | **1** | **2** | **98.0%** | **73** |
- **Cipher words given a value (certain + probable): 146/149 = 98.0%.**
- **Read share, every code counted as a hole: 132/(149 + 73) = 59.5% (certain words only); with the 14 probable words 146/222 = 65.8%.** Under 95%: read in part.
- Held-out set of test 2 (images 81, 83, 84, 87, 90), re-read after the measurement: 50/51 = 98.0% (the 55 units of the
measurement become 51 words once split words are joined).
- Integration recount (12 Nov 2026): image 77 has 22 cipher words, not 23; image 93 has 24 codes, not 26; the code 545 that opens
N. 319 is counted (11 codes on images 78-79). The PR's totals for these letters were 150 words, 133 certain, and 75 codes.
### Probable words (14), one reason each
- 78: PACQUET (19 for 17; the gloss of the time is faint there); HIEMMELIGEN (19 for 10); BORT (14 for 17; the gloss of the time
reads bo(r)t); PREIUDICERLIG (4 for 8); TIENER (17 for 7); IMPORTANCE (n omitted by the encipherer).
- 79 r.: INSTRUCTION (21 taken as one number; tick ambiguous).
- 81: COURTISANER (tick under the 2/6 junction; 14 read for 17, 4 and 7 alike in this hand).
- 83 r.: PROCURATOR (17 written like 14, twice; blot).
- 87: BREFWE (19 for 20; second piece with its own opener 32).
- 90: MED (8 for 7).
- 91: WEN (no tick between 3 and 24).
- 93: SKAL (the 21 has the shape of a u, also read 4); PASSERE (22 clearly written where 3 = a is expected).
### Restored words (1, not counted)
- 83 l.: [hende] (read clearly "hide", 1.11.12.7.8; the sense wants "aff hende", of her, the Queen?).
### Unresolved words (2)
- 78: 31.34.26.13.20 ("bømt"; read clearly and glossed "°m t" at the time; sense not established).
- 93: 30.?.9.2.14=5.6.20 (blot after the opener; "handet").
## What the letters say about the cipher, in clear
- Image 62 (11 Jan 1649, an old-cipher letter): the new alphabet ("dett ny Alphabet") will be sent.
- Image 80 (end of N. 334, 6 Jan 1653): two numbers are to be added to the common nomenclator ("sub N° 349" and "N° 561"); Juel
writes his numbers in letters ("dine bogstaffuer ... denotere tallet") and is asked to write t and k more clearly.
- Image 93 (P.S., 1655): the last cipher sent by Erich Krag had no code for the States General nor for the Øresund.
## Why the holes
The cipher words in letters read to 98.0%. What does not read are the 73 nomenclator codes (49 different numbers): no list of
them is in the unit, the interlinear gloss of N. 319 covers letters only, and only two codes are defined in clear (349 and 561,
on image 80, names read as probable only). Some invite a guess from the sense (770 Sweden? 222 Queen Christina?), but a guess is
not a value, so they stay holes.Translation
# vibejuel1651 - English translation of the cipher passages in the new alphabet (Vibe to Juel, 1651-1655)
Translation of `reading.md`, letter by letter (the contributor's, PR #410). Clear and cipher words are translated together.
Marks: #211 = nomenclator code, not identified (left as a number); [x] = restored word; {x} = probable word; <...> = unresolved
word; ( ) = gloss or doubt. Guesses at the codes are given only as such, in the notes; none is part of the reading.
**Image 77, N. 316.** To have the seal (signet) of #876 rightly engraved, #355 must press the ordinary seal of #876 in wax two or
three times, clearly; then another can be made exactly like the first, so that all stays unnoticed.
**Images 78-79, N. 319.** Some time ago #545 debauched the boy of #818, who took a big {packet} of letters; #818 wrested it from
him and carried it {away} {secretly}, although there was nothing in it that could be {prejudicial} to him. Perhaps there were some
of #873's and #479's letters. The impudent #485 took his leave with the valet of #254, to do his best to debauch my former
{servant} Charles, to see if he could procure some letters of #872. Quo poterant ibant, quo non poterant ibi stabant. Work is now
going on to intercept some letters of #872 of {importance}, for which the wedding, or the Diet (Herredag) <...>, will give an
occasion ... so that one may find some cause to take his estates in #73.
**Image 79 right, N. 334, 6 Jan 1653.** #353 leaves with the first occasion for #770; #397 likewise with #353. I will say nothing
here of their {instruction}, since it will be communicated to you; it can be dealt with at a border meeting (grentze møde) ...
negotiated here, so everything is referred there.
**Image 81, N. 332.** #876 and the young lady (frøken) Hedwig, whether they too were among the {courtiers} who have done ...
**Image 83 left.** Würtz #623, creature and confidant of #376 #357, will persuade #454 that #770 has resolved #463 against #217. It
is said for certain that he has lately been ennobled by #222 in #770 and has received estates from [her]; whereupon one does not
know how far one may trust him here.
**Image 83 right, N. 331.** ... which informs me of Würtz's (#620) credit and esteem in Sweden. Meanwhile I only let you know that
#872 (sends) a {procurator}, #95, with a recommendation to #583.
**Image 84, N. 333.** Since #222 has so many fireships (brandt) #791 built, it would be needful to learn by what name or sign each
of these fireships can be told from the other ships; for there is no doubt they will be fitted out like the other ships.
**Image 87, N. 335.** I will not fail to let you know that {letters} by the new post have come to her from #872 ... what is found
... your and #872's relation ... #872 writes that #601 #397 shall have ... copies of ... letters of #872 ... that #397 ... that no
act was given in writing.
**Image 90, N. 336, 13 Sep 1654.** Let me know what is decided {with} him according to the order that #397 has received about it;
and also what the letter of #879 that was opened contained.
**Image 91, N. 143 / 321, 1654.** I beg and urge you, by our trusted friendship, to let me know who it is that #879 trusts most:
is he the worst, and the best {friend} of #872? I swear (in French) on the salvation of my soul to keep it secret, and I wish to
know it only to manage my friendships hereafter.
**Image 93, N. 167 / 337, 6 May 1655.** ... we are told for certain that #1590 has a design on #1274 <...> #1195 #671, which makes
the States General somewhat perplexed for #430 #889 #428 #487 #770, and they begin to #467 how this #1588 #978 is to be prevented.
#797 #467, if one #798 {must} {pass} them, to #428 #126 #1135 #1590 #1771 by water and lie #448 #600 in the Sound.
Notes
- Codes that the sense invites guesses for (not part of the reading): 770 Sweden (it stands where Sweden is meant, next to
"Swerrig" in clear on image 83)? 222 Queen Christina ("hende", her)? These are the contributor's guesses; no key supports them.
- "Herredag" is the Danish diet/court of the realm; "brandt" = brander, fireship.
- The letters of 1649-1651 in the old cipher are translated in `targets/vibejuel1649/translation.md`.Key: key.tsv
06 Sources
- Rigsarkivet, Copenhagen, Tyske Kancelli, Udenrigske Afdeling (TKUA), Speciel del, Sverige: Resident Peder Juels og Mogens Høegs gesandtskabsarkiver, Ordrer og breve m.m. 1648–55 (leaves stamped “Ny kongl. Saml. Fol. 1299 e”). Arkivalieronline unit 261583, images 77–93: image 77, image 78, image 83.
- Transcription, key, protocols, measurements, reading and translation: Brice Dudout (Europe chiffrée) and his AI agents, pull request #410, 11 November 2026. The protocols, hashes, scripts and held-out transcriptions of both tests are in
frozen/in the repository folder. - Lead: catalogue entry 2527, from the project’s paging of the TKUA envoys’ archives on 11 November 2026, which found the glossed letter N. 319.
- Fryxell, Handlingar rörande Sverges historia (1836), prints a Swedish abstract of Juel’s register of outgoing letters, without the cipher passages (from the contributor’s notes; not opened at integration).
Transcription, key, decoder, frozen tests, reading, translation and notes: targets/vibejuel1651/.
Cite this page
Brice Dudout and Daniel Bourdeau, “Peder Vibe to Peder Juel, 1651–1655: the new alphabet”, Cyphersolver, posted 12 November 2026, https://cyphersolver.com/vibejuel1651.html.