- Sources: primary, discussion
- Summary: The post is a guest post on anthropic.com written by Matt von Hippel, a science writer, with an addendum by Lance Dixon, and it states that the Fable 5.1 model, running inside the paid Claude Science harness, carried out a nine-loop amplitude calculation in N=4 super-Yang-Mills theory, and that Dixon at SLAC validated the result. The post puts the cost at roughly one or two thousand dollars for an end user, of which the bootstrap step took about $100, corresponding to 96 CPUs running for a week. It reports that Song He's group at the Chinese Academy of Sciences had already obtained the majority of the result with GPT-6 assistance, links that concurrent result, and concludes that the model did something humans turned out to be able to do as well.
- Why it matters: The post states that Dixon validated the result independently, and it also states that Anthropic invited and compensated the author, that its staff gave feedback on drafts, and that Dixon received Claude usage credits for the validation, so the vendor funded both the result and its check.
- Follow-up: Track the peer-reviewed writeup of the amplitude result.
send feedback on this story