एआई
Kathmandu, October 21.
A wartime secret message left by Napoleon Bonaparte, which remained an unsolved mystery for over two years, has been successfully deciphered in a matter of hours with the assistance of artificial intelligence (AI). This breakthrough demonstrates that uncovering hidden secrets in historical records is no longer restricted to scholars and experienced cryptanalysts alone.
The message, consisting of two dozen lines featuring a combination of symbols, numbers, and letters, was originally sent to a French general in 1809. Long overlooked in online educational archives, the ciphertext was finally solved by American AI engineer Carter Church.
Using OpenAI’s latest model, 27-year-old Church cracked the code in just six hours. He noted that doing the same work through traditional methods could have taken weeks.
Church, who works as a cybersecurity professional at the American firm SentinelOne, stated that he did not require extensive knowledge of code-breaking techniques or the history of that era. After a few evenings of effort, he consulted Japanese patent engineer Satoshi Tomokio, who maintains an online registry of unsolved historical cryptograms.
According to Church, AI’s primary strength lies in its capacity for continuous, tireless effort. Once the secret of the message was unlocked, he experienced a profound sense of satisfaction and excitement.
Writing on his website, “Cryptiana,” Tomokio noted that 2026 will be remembered as a decisive year for solving historical ciphers. According to him, the majority of the cryptograms on his registry have been successfully solved with AI assistance since mid-September.
Researchers had previously been developing AI models specifically designed to decode historical ciphers. However, the number of solved messages has increased dramatically following the public release of powerful general-purpose AI tools. Church utilized OpenAI’s GPT-6 Astra, while other researchers have employed tools from the American company Anthropic.
Tomokio mentioned that some of the more difficult cryptograms will likely remain mysteries. Due to a scarcity of useful data for decoding, cracking those specific ciphers may continue to pose a challenge in the coming years.
While he is excited about long-standing problems being resolved so easily, Tomokio admitted to feeling mixed emotions upon seeing AI achieve this with such ease. He recounted how a fellow cryptanalyst remarked that AI made them feel redundant.
At a time when AI is outperforming human capabilities across fields ranging from mathematics to computer programming, concerns have also grown regarding changing employment landscapes and the potential displacement of certain jobs. However, the technology has unlocked unprecedented possibilities for historical research. From deciphering ancient documents to analyzing millions of medieval manuscripts and reading ancient texts, AI is proving remarkably useful.
In Italy, researchers are using X-rays and AI to study the Herculaneum papyri—ancient scrolls that were charred and turned into carbon during the eruption of Mount Vesuvius in 79 AD. In June, researchers announced that they had successfully and entirely virtually unrolled one of the hundreds of scrolls that were too fragile to be opened by hand, confirming that it contained a philosophical text on ethics.
Similarly, AI is being deployed to copy millions of pages of Hebrew manuscripts to create a massive, searchable digital archive encompassing rare sources of Jewish history and culture. According to Daniel Stökl Ben Ezra, a project director at École Pratique des Hautes Études in Paris, such a massive volume of data could never be processed by humans alone.
He compared the resulting digital archive of data to a “massive Facebook of the past,” through which researchers hope to uncover connections between people, places, and ideas throughout history.
In Paris, France, an AI system has also been trained to transcribe 14,600 Latin pages of medieval records at the Notre-Dame Cathedral. This project has aided in studying complex historical archives similarly to the deciphering of old ciphers.
According to historian Julie Claustre, who is leading the project, these records are administrative documents hastily written in shorthand by medieval notaries, featuring specialized terminology and complex writing styles.
The AI system achieved an accuracy rate of 88 to 94 percent in character recognition. This breakthrough has yielded valuable insights into the cathedral’s daily life, construction work, employees, and the care of abandoned children.
Claustre noted that transcribing such archives manually would have taken many decades. Furthermore, she emphasized that the importance of AI extends beyond merely saving time; it is fundamentally transforming how history is studied by enabling research into questions that were previously impossible to explore.
RSS/Unified Nepal
