Anthropic Scion post now credits Meta's Mononoke

Anthropic has updated its post about Scion, its in-house source control system, to credit Mononoke, a server that Meta built for its own source control. The change came on October 15, 2026, a day after a developer said Scion's design closely copies Mononoke's without saying so. The dispute is about credit, not code: the developer says copying the ideas is fine, but hiding where they came from is not.

What Anthropic announced

Anthropic described Scion on October 13. We covered how it replaced git for Anthropic's main repository after daily merges passed 30,000. Scion keeps history on a server, and each machine fetches only the files it needs.

One idea in the post is "derived data." These are extra indexes that the server builds from each change, so that slow operations such as blame become fast. Blame is the tool that shows who last changed each line of a file. Anthropic's post said the design means "we can easily add a new type of derived data when a new data structure would make some operation faster."

The complaint

On October 14, a developer who writes as Rain on the sunshowers blog said that design matches Mononoke. Mononoke is the server part of Sapling, the source control system Meta uses. Rain says they described Mononoke's derived-data design in writing in 2022.

At the time, Anthropic's post credited Sapling only for "hard-won lessons," Rain wrote, and did not name Mononoke. "I think the level of imitation goes well beyond having similar insights," Rain wrote, calling it close to plagiarism. Rain added that they had no objection to the design being copied.

Rain also asked two AI models about prior art for Scion's design. Both Anthropic's Opus 5.5 and OpenAI's GPT-6 Astra pointed to Mononoke, according to the post.

The follow-up post

Rain published a second post on October 15 and went further. Rain says private sources confirmed that Scion copied Mononoke on purpose, and that some of the same people worked on both. In the first post, Rain had called the likeness mostly unintentional.

The second post says the resemblance covers Scion's whole data model. Scion's changesets, the records of each change, match what Mononoke calls Bonsai changesets, according to Rain. Rain says Bonsai was an unusual design when it was created in 2018. Rain also contrasts the timelines: about 10 years of work at Meta, against the five months Anthropic took to build Scion.

"The problem is attribution, not copying," Rain wrote. These are Rain's claims; Anthropic has not addressed them point by point in its post.

What Anthropic changed

The updated Anthropic post now carries a note under its byline. "Updated October 15 to include more detail on the projects that pioneered many of the techniques we use in Scion," it reads.

The body now says that source control for huge monorepos "has prior art, notably Google's Piper and Meta's Sapling." Mononoke is named in a footnote: "Specifically, the server component of Sapling, Mononoke." The post's byline lists Nick G. and the source control team as authors.

Rain's second post says an Anthropic engineer, Nick G., responded on Hacker News and that the credit was then added. Rain says they stand by both posts.

What this means for developers

Credit the part, not just the project. Anthropic had already named Sapling, yet the people who built Mononoke still felt erased. If your design borrows from a specific component, name that component in the body of your post, with a link. A footnote fixes the record, but readers skim past footnotes.

AI tools make this easier to get right and easier to get wrong. An agent can study an existing system and help rebuild it in months, as Anthropic's own five-month timeline shows. Keep a short design note listing every system you studied, and check each one's license before any code moves across.

If you are weighing Scion's ideas for your own monorepo, the update is also a reading list. Anthropic will not release Scion, but its post now points to Piper, Sapling and Mononoke as the earlier work. Those are the systems to study, along with their authors' write-ups.

The next thing to watch is whether Anthropic says more. So far its only public answer is the update note and the new footnote.