GPT-6 Astra 六小时破译拿破仑 217 年前密信
一张模糊照片、一句提示词,Astra 读完 1300 个符号并找到旧密钥,解出 1809 年军事简报,还纠正了史学界记错的年份。


Napoleon's 217-year-old mysterious letter cracked by AI in 6 hours it was written in 1809, full of symbols nobody could read, sent by napoleon's orders to one of his generals. two centuries, and nobody ever cracked it a cybersecurity engineer gave gpt-6 astra a single blurry photo of the letter and one prompt. the model read all 1,300 symbols, found an old forgotten key that connected to this exact letter, and used it to decode the whole thing turns out it was a military briefing, troop positions across europe, sent two weeks before austria got invaded. astra even corrected the date historians had wrong, moving it from 1807 to march 1809 what's wild isn't that it got decoded, people probably could've done that eventually. it's that one model did the entire thing start to finish in 6 hours, reading the image, cracking the code, checking it against real history, with just one prompt there's probably thousands of letters like this sitting in archives that nobody ever read because it wasn't worth a historian's years. that just changed now