More on Anthropic's Scion ripping off Mononoke

The follow-up, dated October 15, 2026, comes from a former Mononoke contributor who goes beyond the first post's "derived data" argument. The author says Scion's changesets match Mononoke's Bonsai changesets almost exactly, and that the name reverses the syllables of Bonsai. The author also says private confirmation after the first post showed the copying was intentional and that personnel overlapped. That claim is the author's own and has not been independently verified. Anthropic's October 13 announcement cited Meta's Sapling as prior art but did not name Mononoke. The post now carries an update saying an Anthropic engineer responded on Hacker News and that sufficient attribution has been added to the blog post. The author still stands by both posts as written. The core objection is that the story of an LLM rebuilding a complex system in five months looks different when the system was closely modeled on a decade-old one. Most commenters still think the Sapling credit was a dodge, while a minority see two large companies squabbling. For founders, the lesson is practical. LLMs tend to reproduce open-source designs that sit in their training data. Teams that publish "built by AI" engineering stories should record which prior systems the model was pointed at and credit them specifically. Leaving that out invites a public correction and weakens the capability claim.