Anthropic unveiled Scion, a new version control system, in a post published on October 13, 2026, describing how it uses a concept called "derived data" to speed up operations such as blames. Within a day, the announcement was being compared to a system built at Meta almost a decade earlier.
In a post on sunshowers.io dated October 14, an engineer who says he helped build Meta's Mononoke source control server argued that the similarities go well beyond shared technical insight. He points to his own 2022 Oxide candidate materials, where he described using "derived data" for exactly the same purpose, and calls the level of imitation close to plagiarism.
Mononoke, Sapling and the credit question
The engineer's main complaint is not the copying itself but the attribution. He notes that Sapling, another Meta system, was credited only for "hard-won lessons", while Mononoke was not mentioned at all. In his reading, the post leaves the impression that Claude designed the architecture, when his team at Meta developed many of those insights years earlier.
He also points out that Mononoke was not built in a vacuum: his team studied prior art ranging from Google's SrcFS to write-ahead logs in databases. Mononoke is open source, he adds, and that is deliberate — releasing it was meant to spread good ideas, and he says he has no problem with Anthropic reusing the design.
Why AI credit matters now
The sharper warning in the post concerns how AI systems are trained. Mononoke sits in the training data of large language models, so when those models help design new systems they tend to converge on designs that already exist.
As evidence, he says that asking Opus 5.5 or GPT-6 Astra about prior art immediately surfaces the similarities to Mononoke. His conclusion is blunt: with models now able to trace influences so easily, giving credit to earlier work is easier and more important than ever. He also points readers to a blog post by Vicent Marti for Cursor as an example of a system that shares some insights but makes key design choices differently.
