A former Meta engineer who helped build Mononoke, an open-source source control server, says Anthropic’s new version control system borrows heavily from that project and fails to give it credit. The engineer says Anthropic’s description of Scion, a system the company published this week, uses the same “derived data” concept he described in 2022 as part of his materials for a job application at Oxide.
In Anthropic’s post, the company says Scion uses derived data to make operations like blame more efficient. The engineer posted side-by-side screenshots of his earlier description of Mononoke, which uses the same term for the same purpose. He argues the similarities go beyond a few shared ideas.
“I think the level of imitation goes well beyond having similar insights,” he wrote. He said the post credits Sapling only for “hard-won lessons” and does not mention Mononoke at all. He called that omission “borders on plagiarism.”
He says the post suggests Claude came up with the design on its own. In his account, the insights were developed by his team at Meta roughly a decade earlier, and Claude later applied them. His team, he wrote, also looked at prior art for both Mononoke and Sapling, including Google’s SrcFS and write-ahead logs used in databases.
He was careful to separate the credit question from the copying itself. He said he is glad the team’s design has held up, and he does not have a problem with Anthropic borrowing it. He noted that Meta released Mononoke as open source specifically so good ideas could spread.
His larger concern is about how AI tools learn. As large language models help engineers design new systems, he argues, they will tend to lean on existing designs. Mononoke is open source and, in his view, part of the training data for every major model. He said that when he asked two models, which he named as Opus 5.5 and GPT-6 Astra, about prior art, both immediately pointed to the similarities with Mononoke.
He closed his post with a call for credit. “Giving credit to those that came before us is now easier — and more important — than ever,” he wrote. “There is no longer an excuse.”
He also pointed to a system he considers a fair comparison. Vicent Marti’s blog post about Cursor describes similar insights, he said, but with many key design differences.
Anthropic’s post has not been updated in response to the criticism, based on the material available for this story. The company did not respond to a request for comment before publication.
Questo articolo è stato scritto con l'assistenza dell'IA.
News Factory APP - notizie agentiche per potenziare il tuo SEO e AEO.