“Dahka”, Cracow (1715)

At a glanceCiphertext-only break · partial (84% of the cipher)

The letter: A Polish letter from Cracow signed “Dahka” (catalogued as Dahku), probably of 21 September 1715, to an unnamed member of the Branicki circle; 17 lines of graphic signs on pp. 2–3, with clear words among them.
Where: AGAD Warsaw, fonds 338 (Zbiór Anny z Potockich Branickiej), unit 1/338/0/1/682, scan 3. The archive card says “część listu pisana szyfrem, nierozwiązanym”. Not on DECODE.
The cipher: A letter-for-letter substitution of Polish with 24 signs for 20 letters and a few homophones. A framed sign and a tied sign stand where names are expected.
How it was deciphered: Broken on the cipher alone, with no crib, by annealing with a Polish model learnt from the Deboli clear texts of this repository; eight restarts gave the same text. Measured once on 72 held-out signs: 0.444 against 0.083 for shuffled keys, under the 0.60 threshold, so the key is partial.
What it says: The Sandomierz deputies asked that the Crown army support the levy of their voivodeship, set for 9 October, and not be moved away from Sandomierz; the answer was that nothing could be promised before the Primate’s resolution was known.
Still open: Seven words that do not construe, two signs that are probably names, and a word at the edge of the page.
Credit: Brice Dudout (Europe chiffrée) transcribed the signs, recovered the key and read the passage, working with Claude Opus 5.5 agents. The letter came from Daniel Bourdeau’s catalogue (no. 1581).

01 The letter

The letter is in the papers of Anna Branicka, née Potocka, at the Central Archives of Historical Records in Warsaw. Page 1, the address and the end are in clear Polish; only scan 3 (pp. 2–3) carries cipher. No gloss, key or clear copy is on the five scans. The unit card gives “1715, IX, 21”, and the cipher speaks of 9 October as still to come, so the letter is probably of 21 September 1715. That autumn the nobility of the Crown was arming against the Saxon troops of Augustus II, the movement that became the Confederation of Tarnogród in November.

02 The key and its test

The signs were named with ASCII letters in a table fixed before the line-by-line transcription, and the file was frozen before any attempt at the key. Lines d01–e02 (278 signs) were the training part: simulated annealing with a Polish four-letter model, eight restarts, the same text each time. The raw key was frozen before the held-out lines e03–e06, chosen before any attack, were decoded.

The measure is the share of decoded signs that fall in Polish words of five letters or more: 0.444 on the held-out block, against 0.083 for the best of 200 shuffled keys. The margin is met, the 0.60 threshold is not; the block is short and there is no clear text of the time to compare with. After the measurement a blind shape question settled the barred sign V as W in two places, for the reading only.

03 The reading

63 cipher words: 39 certain, 14 probable, none restored and 10 unresolved, 84.1% read by sense. The deputies of Sandomierz asked for the Crown army (wojsko koronne) to support the levy en masse of the voivodeship, set “pro die nona octobris”, and not to be moved away from Sandomierz. The answer was that no commitment could be made before the Primate’s resolution was known; it was only declared that the army would not be withdrawn but taken across the Vistula. The Polish and an English translation are under Full texts.

04 What remains uncertain

Seven words do not construe as letters: zto, e, kostalacio (d02), kodzarzyc (e05, where a place name is expected), nadra, zo, a (e05–e06).
The framed sign (twice) and the tied sign (once) are probably names, such as the subject of “chciał assistere” and a title before “prymas”; no nomenclator for this cipher is known.
The last word of e05 was transcribed from a reduced view at the edge of the page.
The key cannot be tested again on this letter; another text in the same signs is needed.
The writer and the addressee are not identified.

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.

scan 3, pp. 2-3: the cipher passage of the letter (17 lines) read in part

The Sandomierz deputies asked that the Crown army support the levée en masse of the Sandomierz voivodeship, set for 9 October, and not be moved away from Sandomierz; the answer was that no commitment could be made before the Primate's resolution to the deputies was known; it was only declared that the army would not be withdrawn but taken across the Vistula. The run-up to the Confederation of Tarnogród (Nov 1715).

Builds coherent sentences, but under the read bar of 95% of words read as sense (84% here); the holes are marked in the reading.

Transcription

transcription.txt

# dahku1715 - AGAD Warsaw, fonds 338 (Zbior Anny z Potockich Branickiej), 1/338/0/1/682, scan 3 (pp. 2-3):
# letter signed 'Dahka' (catalogue 'Dahku'), Cracow, 1715. One AI agent transcribed the signs, 5 Nov 2026, with a sign table fixed
# before the line-by-line transcription; the file was hashed (frozen) before any attempt at the key.
# One ASCII letter per sign shape (no value assumed): 8 4 2 o + x f q l T V C N j F z P W a b B n c #; =word = a clear Polish word
# written among the cipher (spelling of the letter, no diacritics), used only as a word break, never as known plaintext.
# d = p. 2 (left page), e = p. 3 (right page). Lines e03-e06 = held-out block (never decoded before the single measurement).
# Superscripts not transcribed: d05 '2' over the first b; d07 a P-like sign over the end; e01 a 'g'-like sign over C.
# Doubt: e05 last word (page edge) read on the reduced view. After the measurement we noticed that the isolated '+' before
# '=domieszky' in e06 is probably the clear word 'to' (+o); not corrected here (frozen file). No images here.
d01	=u =Jmd =krakowskiey =WD jBaTBbqn anc
d02	FBfq8a4l W2B n C xBa2PT+4qB =aby
d03	N 4#4q+T +aaqa2n8n WbBqaxqf
d04	xB8Bcclf jBajBT2nfP 8PaWncqP
d05	bF2b+ ancFBfq8axqnVB =ktore Fn2n8
d06	fqcBb+cB j8B Fqn cBc+ B42Bo8qa
d07	=y =zeby bBqax+ xB8BccnVB cqnBFF+T+
d08	BFancFBfq8W+ c+2B BFcqnaTq 2+
d09	xq 8najBca =ze =deklarowac =y =tym N
d10	cqnfBWn b+aaqa2n4l WbBzaxqf
d11	=potewaz bj8WBF 28Wno+ bqnFWqn4
e01	=iaky 8naBTqP4qB F+ C qf j8zf+a
e02	jBaTBf ancFBfq8axqf =domey =posl
e03	=nym bBqaxB W+a cqnBFF+Tq+4 F+
e04	xT+8Bb+T =tylko W+bqaTnc j8Wn
e05	j8+bq4 W xBFVP8Wl4 =prosto c+F8P
e06	WB a28Bcn + =domieszky =deklarowac

Decryption

decryption.txt

# dahku1715 - mechanical decryption of transcription.txt with key.tsv (one value per sign), no correction.
# =word = clear word of the letter, kept as written. Line references as in transcription.txt.
d01	=u =Jmd =krakowskiey =WD poslowie sen
d02	domirscy zto e w kostalacio =aby
d03	z chcial assistere zwoiskim
d04	koronnym pospoltema raszenia
d05	wdtwa sendomirskiego =ktore deter
d06	minowano pro die nona octobris
d07	=y =zeby woiska koronnego nieoddala
d08	odsendomirza nato odniesli ta
d09	ki respons =ze =deklarowac =y =tym z
d10	niemoze wassistecy zwoyskim
d11	=potewaz wprzod trzeba wiedziec
e01	=iaky resoliacio da w im prymas
e02	poslom sendomirskim =domey =posl
e03	=nym woisko zas nieoddaliac da
e04	klarowal =tylko zawislen prze
e05	prawic z kodgarzyc =prosto nadra
e06	zo strone a =domieszky =deklarowac

Reading

reading.md

# dahku1715 - reading of the cipher passage (Polish, spelling of the letter)

Marks: plain = certain (every letter given by the key); {x} = probable (one or two signs off, or a letter the key does
not give, settled by the sense); [x] = restored letters inside a clear word (no cipher word is restored); ‹x› = unresolved
(the decryption as it stands). Clear words of the letter are kept as written. Line references as in transcription.txt.

The key is frozen and partial by our protocol (see validation.txt). After the measurement Brice Dudout answered a blind
shape plate on the barred sign `V` (planche V, 5 Nov 2026, 23:41): at d05 and e05 it has the shape of `W` (= z), so it is
read z at those two places, for the reading only (the frozen key and the measurement are unchanged; d07 was not asked and
keeps V = g).

| line | decryption (key.tsv, V = z at d05 and e05) | reading |
|---|---|---|
| d01 | U JMD KRAKOWSKIEY WD poslowie sen- | u JMD krakowskiey W.D. : posłowie sen- |
| d02 | domirscy zto e w kostalacio ABY | -domirscy ‹zto› ‹e› {w} ‹kostalacio› aby |
| d03 | z chcial assistere zwoiskim | ‹N› chciał assistere z woiskim |
| d04 | koronnym pospoltema raszenia | koronnym {pospolitego} {ruszenia} |
| d05 | wdtwa sendomirskiezo KTORE deter- | {wdztwa} {sendomirskiego} które deter- |
| d06 | minowano pro die nona octobris | -minowano pro die nona Octobris, |
| d07 | Y ZEBY woiska koronnego nieoddala | y żeby woiska koronnego {nie} {oddalać} |
| d08 | odsendomirza nato odniesli ta- | od Sendomirza. Na to odnieśli ta- |
| d09 | ki respons ZE DEKLAROWAC Y TYM z | -ki respons, że deklarować y tym ‹N› |
| d10 | niemoze wassistecy zwoyskim | nie może {w} {assistencyi} z woyskim, |
| d11 | POTEWAZ wprzod trzeba wiedziec | ponieważ wprzód trzeba wiedzieć, |
| e01 | IAKY resoliacio da w im prymas | iaky {rezolucyą} da ‹C› {im} prymas |
| e02 | poslom sendomirskim DOMEY POSL- | posłom sendomirskim. Domey posł- |
| e03 | NYM woisko zas nieoddaliac da- | -[a]nym woisko zaś nie oddaliać {de-} |
| e04 | klarowal TYLKO zawislen prze- | {-klarował}, tylko {za} {Wisłę} prze- |
| e05 | prawic z kodzarzyc PROSTO nadra | -prawić z ‹kodzarzyc› prosto ‹nadra› |
| e06 | zo strone a DOMIESZKY DEKLAROWAC | ‹zo› stronę ‹a› domieszky deklarować. |

Notes on single words:
- `N` (a framed sign, twice, d03 and d09) decodes z, but in the reading it is the subject of "chciał assistere" and of
  "deklarować nie może": most likely a nomenclator name (the hetman? the king?). Counted unresolved.
- `C` (e01, ".cш." tied, with a г-like superscript) decodes w; it stands where a name or title is expected ("da ‹C› im
  prymas"). Counted unresolved.
- d05 {sendomirskiego}: with V = z the cipher spells "sendomirskiezo"; the sense (wdztwa sendomirskiego) is clear, and the
  same ending is spelled with V in d07 "koronnego". Counted probable, not certain, because one cipher letter disagrees.
- e03-e04 {deklarował}: the cipher has "da klarowal" (a for e). Counted probable.
- e05 ‹kodzarzyc› (V = z; with V = g "kodgarzyc"): a place name is expected after "przeprawić z"; not identified.
- e05-e06 ‹nadra› ‹zo› … stronę: perhaps "na drugą stronę" (to the other side), but the letters do not give it; unresolved.
- e06 ‹a›: after the measurement we noticed that this isolated `+` is probably the clear word "to" (+o); not corrected.
- "posł[a]nym": the letter a is restored inside a clear word (not counted, clear words are outside the count).

## Count (Daniel's rule: certain + probable over all cipher words; restored words would not count)

Cipher words, cut as in the reading above (a cipher token like "zwoiskim" counts as two words, z + woiskim):
**63 words: 39 certain, 14 probable, 0 restored, 10 unresolved -> 53/63 = 84.1% read by sense**, under 95%:
**read in part**.
(Counted on the cipher tokens as written, without splitting z + noun or nie + verb, the same text gives 49 of 59 = 83%.)
The planche V did not change the share: d05 moves from certain to probable, e05 stays unresolved.

## Reading as running text

… u JMD krakowskiey W.D.: posłowie sendomirscy ‹zto› ‹e› {w} ‹kostalacio› aby ‹N› chciał assistere z woiskim koronnym
{pospolitego} {ruszenia} {wdztwa} {sendomirskiego}, które determinowano pro die nona Octobris, y żeby woiska koronnego
{nie oddalać} od Sendomirza. Na to odnieśli taki respons, że deklarować y tym ‹N› nie może {w assistencyi} z woyskim,
ponieważ wprzód trzeba wiedzieć, iaky {rezolucyą} da ‹C› {im} prymas posłom sendomirskim. Domey posł[a]nym woisko zaś nie
oddaliać {deklarował}, tylko {za Wisłę} przeprawić z ‹kodzarzyc› prosto ‹nadra› ‹zo› stronę ‹a› domieszky deklarować.

Translation

translation.md

# dahku1715 - English translation of the cipher passage

Translation of reading.md (Brice Dudout, with a Claude Opus 5.5 agent, 6 Nov 2026). Probable words in {braces},
unresolved stretches as ‹…›. Polish names and titles kept where the sense is uncertain.

> … at His Grace's of Cracow (W.D.): the Sandomierz deputies ‹…› ‹…› {in} ‹…›, [asked] that ‹N› would assist, with the
> Crown army, the {levée en masse} (pospolite ruszenie) {of the voivodeship} {of Sandomierz}, which was set for the ninth
> day of October, and that the Crown army {not be moved away} from Sandomierz. To this they received the answer that ‹N›
> cannot commit himself to this {on assistance} with the army, because one must first know what {resolution} ‹C› the
> Primate will give {them}, the Sandomierz deputies. To the deputies sent home he {declared} that the army would however
> not be moved away, only taken {across the Vistula} from ‹…›, straight ‹…› ‹…› side ‹…›, to declare [this] to those
> at home (?).

Context (to be checked by a historian): autumn 1715, the run-up to the Confederation of Tarnogród (November 1715), when the
Sandomierz nobility rose against the Saxon troops of Augustus II. The unit card dates the letter 21 Sep 1715 ("1715, IX,
21"), and the cipher speaks of 9 October as still to come; our first sorting note (21 May) is probably wrong. The date line
of p. 1 was not re-read for this file.

Key: key.tsv

05 Sources

AGAD Warsaw, 1/338/0/1/682, scans 1–5 (szukajwarchiwach.gov.pl, unit 28224391); not reproduced here.
Key, transcription, reading and validation: targets/dahku1715, contributed by Brice Dudout, with Claude Opus 5.5 agents.

Notes, transcription, decryption, reading, translation and the key are in targets/dahku1715.

Cite this page

Brice Dudout and Daniel Bourdeau, ““Dahka”, Cracow 1715: a sign cipher in a Polish letter”, Cyphersolver, posted 6 November 2026, https://cyphersolver.com/dahku1715.html.

“Dahka”, Cracow, 1715