Developer Prinz announced that ChatGPT 6 Astra successfully deciphered a 108-year-old encrypted German military radio transmission from World War I [2]. Earlier manual efforts failed. Using historical cipher principles paired with naval archives, the artificial intelligence model revealed tactical intelligence detailing Allied deployments off the Crimean Peninsula [1]. Archival verification confirmed that the decoded alert matched logbooks from the British cruiser HMS Canterbury [2].
ChatGPT 6 Astra Tackles Unsolved Wartime Messages
Historical radio intercepts often resist analysis. The Crimean transmission was no exception. German naval communications from late 1918 created specialized puzzles that persisted across generations of cryptographers without resolution [1].
Prinz selected the target cipher from a prominent compilation of 50 unsolved historical cryptograms hosted on the German science portal Scienceblogs.de while writing on their Substack [1]. The catalog features legendary enigmas spanning from the undeciphered Voynich manuscript to encrypted notes associated with serial killers. Among those archival challenges rested a cluster of German military dispatches encoded using the ADFGVX method (a specialized field cipher that combined letter substitution with fractionated transposition). While prominent codebreaking expert George Lasry previously cracked hundreds of related wartime intercepts, roughly a dozen stubbornly resisted every human effort [2].
Addressed to the German High Command and intended for an admiral or Naval Command, the surviving transcript appeared as an indecipherable string of letters. Could modern generative networks discover patterns where classical statistical routines stalled? Prinz tasked ChatGPT 6 Astra with analyzing the encrypted text, allowing the system to test contextual linguistic structures rather than relying exclusively on brute computational force [1]. By evaluating the transmission through historical cryptographic literature, the model systematically isolated candidate arrangements that matched Imperial German communications doctrine [2].
Deconstructing the German ADFGVX Cipher Grid
Frontline secrecy demanded robust encryption. The Imperial German military introduced the ADFGVX cipher in 1918 to protect sensitive tactical communications against Allied radio interception [1]. German signal officers deliberately chose the letters A, D, F, G, V, and X because their Morse representations differed sharply, minimizing transcription mistakes during frontline Morse transmission. In this framework, operators distributed thirty-six characters across a six-by-six coordinate grid containing twenty-six alphabet letters as well as ten numerical digits from zero to nine. Each character translated into a two-letter coordinate pair based on its row and column positions [2].
To demonstrate this coordinate mapping, Prinz illustrated the substitution stage using a sample grid keyed with the word HOUSE. In that demonstration table, the letter H resided at coordinates AA, the letter O mapped to AD, and the letter B corresponded to DA. Under those hypothetical coordinates, encrypting the five-letter name PRINZ produced the intermediate cipher string FX GD DX FV VD. Any shift in the underlying substitution keyword rearranged the thirty-six interior cells entirely, producing wholly different coordinates for identical plaintext characters [2].
Fractionation formed only the initial layer. Fractionated transposition scrambled those coordinate pairs across vertical columns according to a second secret keyword. Historical cryptanalysts required both keys. Prinz identified the necessary historical parameters within the specialized monograph The History and Principles of German Military Ciphers, 1914–1918 authored by military cryptanalyst J. Rives Childs. Pages 214 and 215 of Childs’s study documented known wartime keys, while page 217 recorded the specific unsolved dispatch originally transmitted on November 27, 1918. Feeding these historical coordinates into ChatGPT 6 Astra allowed the model to simulate transposition patterns against known imperial formats [1, 2].
Transposition Mathematics Behind the German Keyword
ChatGPT 6 Astra solved the elusive cipher by applying the German keyword TRUPPENVERSCHIEBUNG (meaning troop movement), a nineteen-letter operational term documented in wartime archives [1, 2]. Deciphering the message required sorting the nineteen letters of TRUPPENVERSCHIEBUNG into alphabetical order, placing T sixteenth and R thirteenth. Operators distributed the 170 encrypted characters into eight rows of nineteen symbols plus a ninth row containing eighteen symbols. This transposition grid formed eighteen columns of nine characters and one eight-character column under letter G. Because T was sixteenth alphabetically, it followed fourteen nine-symbol columns and one eight-symbol column, producing an offset of 134 symbols (calculated as 9 × 14 + 1 × 8). The 135th character was ciphertext letter A [2].
A parallel calculation resolved the next coordinate component. Because the letter R was the thirteenth letter alphabetically, it had eleven nine-symbol columns and one eight-symbol column before it, yielding an offset of 107 symbols calculated by multiplying eleven by nine and adding eight. The 108th symbol in the transmission was ciphertext letter V. Combining these coordinate positions yielded the pair AV, which mapped directly to E in the substitution table. That single decoded letter formed the opening character of the German word EIN [2].
Iterative decoding unlocked the text. ChatGPT 6 Astra executed this unwinding without human intervention, iteratively assembling fragmented letter pairs into coherent German military syntax [2]. What made this computational achievement notable was the model’s capacity to synthesize historical cipher tables with complex wartime linguistic patterns that had baffled conventional decoders for over a century. The model recognized that seemingly random letter clusters conformed to standard imperial communication protocols [1].
Decrypted Warning Matched HMS Canterbury Logs
The complete decipherment yielded a tactical dispatch: “EIN ENGLISCHER KREUZER EINLIEG X SEWASTOPOL X S4STEN X EIN GESCHWADER DER X ALLIIERTEN FOLGT 26STEN X.” Translated into English, the message alerted command: “AN ENGLISH CRUISER ARRIVED AT SEVASTOPOL ON THE ?4TH AN ALLIED SQUADRON FOLLOWS ON THE 26TH.” In German naval communications practice, the letter X functioned as punctuation to separate distinct operational clauses [2]. The corrupted symbol S4STEN represented an apparent transmission typo for 24STEN, caused by a frontline telegraphist striking an erroneous key during transmission [1].
To corroborate the finding, ChatGPT 6 Astra cross-referenced the decoded claims against historical naval records. British Admiralty archives confirmed that the light cruiser HMS Canterbury entered Sevastopol harbor in Crimea on November 24, 1918 [1]. Furthermore, official logs recorded that an Allied naval squadron followed into the Crimean port on November 26, matching the decrypted transmission precisely. Much like archival research that documented an elusive mammal rediscovered after six decades of presumed absence, resolving forgotten historical episodes requires validating neglected source documents against physical field records [2].
Archival logbooks verified the tactical warning. The British cruiser HMS Canterbury docked at Sevastopol on November 24, 1918, exactly as the decrypted radio alert indicated [1].
The alert carried immediate strategic weight for imperial forces near the Crimean Peninsula. Addressed to the German High Command and marked for an admiral or Naval Command, the dispatch alerted German units that an English cruiser and an Allied squadron were entering Crimean waters. Had Allied forces intercepted the communication at the time, it would have shown that German signal stations still monitored opposing warship movements despite collapsing frontline logistics during late 1918 [1].
Calendar Assumptions Kept the Message Concealed
Why did this single radio message remain unbroken for over a century? A calendar assumption caused delay. ChatGPT 6 Astra hypothesized that earlier cryptanalytic attempts failed because researchers operated under a flawed historical assumption regarding German keyword schedules. Cryptanalysts previously believed that German signal units only adopted the keyword TRUPPENVERSCHIEBUNG beginning on December 9, 1918, which led specialists to dismiss the phrase for earlier autumn dispatches [1]. However, historical records in Childs’s treatise confirmed that this particular transmission occurred earlier, on November 27, 1918. While Tom’s Hardware noted the dispatch date as November 29, both dates predated the December keyword registry and prevented previous researchers from testing the nineteen-letter key [1, 2].
Online commentary across technical communities reflected on this missed connection. Reader discussions on Tom’s Hardware debated why previous code sleuths never ran all known keywords across every intercept, prompting reader Penfolduk to note that computational resources remained severely constrained for most of the twentieth century. Testing exhaustive permutations on short tactical fragments held minimal priority when the messages appeared unlikely to alter military outcomes. Technology editor Mark Tyson noted for Tom’s Hardware that the successful decoding by GPT-Astra, also discussed online as Astra GTP, provided a clear demonstration of flexible problem-solving on constrained historical datasets [1].
Over a dozen wartime ADFGVX dispatches remain completely unsolved on Scienceblogs.de. Unsolved wartime riddles still endure. Developer Prinz demonstrated that artificial intelligence models like ChatGPT 6 Astra can re-examine neglected historical parameters to unlock century-old transmissions without requiring novel cryptographic theory [1, 2]. Whether modern neural architectures can extend this method to resolve the remaining wartime radio messages depends on recovering additional historical keyword registries from surviving European archives [1].
- ONLINE NEWS Tyson, M. (2026, September 19). ChatGPT-6 Astra cracks 108-year-old unsolved WWI German code for the first time — radio message sharing enemy movement intelligence had evaded decoding, 1918 Crimean fleet warning verified against HMS Canterbury logs. Tom’s Hardware. [Article Link]
- WEBSITE Prinz. (2026, September 17). GPT-6 Astra Solves a WWI German Radio Cipher. Prinz. [Article Link]
