GPT-6 Astra2026-10-02 01:39:11GPT-6 Astra cracks a 217-year-old Napoleon cipher in about six hoursA 217-year-old encrypted letter sent by Napoleon in 1809 has been deciphered with the help of GPT-6 Astra, according to the source material provided. The document had long appeared on Cryptiana’s list of unsolved historical ciphers and survived only as a low-resolution black-and-white reproduction published in a 1969 French military magazine. The image measured 1202 × 1836 pixels and contained roughly 1,300 hand-drawn symbols, later grouped into 155 distinct units. The effort was led by Carter Church, a senior AI engineer at SentinelOne rather than a professional historian or cryptanalyst. The source says he gave Astra a blurry image and a short prompt, after which the model spent about six hours transcribing the symbols, identifying partial anchors from historical material, writing code based on simulated annealing, and validating the output against 19th-century French language patterns. The decoded text reportedly revealed a military briefing sent to Marshal Marmont and restored a phrase that had remained incomplete for 160 years in the published correspondence of Napoleon. Cryptiana operator Satoshi Tomokiyo later said the case was solved after reviewing a code package with validation scripts.20
AI Security2026-08-25 13:22:08AI security startup Alice raises $140 million led by Apax DigitalAI security startup Alice has raised $140 million in a new funding round, according to Bloomberg. The round was led by Apax Digital, a fund under private equity firm Apax Partners, with participation from cybersecurity company SentinelOne and Samsung Electronics. Following the financing, Alice’s valuation is now close to $1 billion. The company focuses on securing AI models and helping enterprises reduce the risk of AI systems being misused, attacked, or behaving abnormally. Alice is already working with major companies in the AI sector, including Anthropic and Google. The deal adds fresh capital to a company operating in one of the more closely watched areas of AI infrastructure: security controls designed to protect models and enterprise deployments as adoption expands.780
