sunshowers · 2026-10-14 · notable
Rain — Anthropic's Scion leans on Meta's Mononoke without credit
Rain, who helped build Meta's Mononoke source control server, argues that Anthropic's new Scion system reuses Mononoke's derived-data design. They say Anthropic's post never names Mononoke, and call the omission close to plagiarism.
A Mononoke builder says Scion copies Meta's derived-data design and asks AI-era engineering posts to credit prior art.
What is it?
Rain's post compares Anthropic's Scion, its new source control system, with Mononoke, the Meta server they helped build. Both pair minimal changesets with "derived data": extra indexes such as blame and path history, built from a commit's parents. Rain says Anthropic credits Sapling only for "hard-won lessons" and never mentions Mononoke.
How does it work?
The argument lines up Anthropic's Scion description against Rain's 2022 write-up of Mononoke. Both define each index recursively from parent changesets, make blame scale with the commits that touch a file, and hide internals from clients so new indexes need no client change. Rain also reports that Opus 5.5 and GPT-6 Astra both named Mononoke when asked about prior art.
Why does it matter?
Rain argues that when LLMs help design systems, they will lean on open-source designs they saw in training, so crediting predecessors matters more than before. They do not object to Anthropic reusing the design, only to the missing credit: "There is no longer an excuse."
Who is it for?
infrastructure engineers and anyone following AI-built systems