Historian puts Opus 5.5 and GPT‑6 on 17th‑century alchemical papers
Benjamin Breen said Opus 5.5 read 5,000 primary sources on Newton's alchemy, and that two of the ciphers he set it had already been cracked.
Benjamin Breen, a historian, published an account on Thursday of putting Claude Opus 5.5 and GPT-6 on early modern manuscripts. He said one of four runs pointed at a source for Isaac Newton's alchemical reading.
Breen said Opus 5.5 downloaded more than 5,000 primary source files from the Samuel Hartlib papers and cross-referenced them across several languages. It found that Newton and Hartlib each used a different anagram for vitriolum, Hungarian vitriol, which Breen said suggests Newton drew on Hartlib's manuscript.
Two of the four experiments landed on work already done. Breen said Opus 5.5 partly deciphered encoded letters of Charles V, then found that cryptographers had cracked one of them in the 1530s and the other in 1916.
On John Dee's Liber Loagaeth, a manuscript long treated as coded, Breen said GPT-6 Astra concluded the text is almost entirely nonsense syllables. The model identified one passage encoding Bornogo, an angelic figure from Dee's own mythology.
A fourth run, on the Darwin Correspondence Project, has produced no decisive result yet, Breen said. His post drew 48 points and 9 comments on the Hacker News front page. No other historian has checked the findings.