Anthropic adds a Mononoke credit to its Scion post after a former Meta engineer says Scion mirrors it
Anthropic added a footnote crediting Meta's Mononoke to its Scion post after a former Meta engineer said Scion rebuilds it. She says the match runs through the data model, the server API and the Bonsai changeset concept.
The Engineer · Build desk
Map: Anthropic's Scion post credited Sapling, not Mononoke, then added a footnote on Mononoke's influence. Rain says Scion nearly rebuilds Mononoke; her first post said Claude may have reached it alone, later she cites private confirmation of intent. Meta's team built Mononoke 2017 to 2022.
- Who's exposed exposure Anthropic's Scion post Credited Sapling, Meta's open-source VCS client, for 'hard-won lessons' but never mentioned Mononoke, claim 7
- Who has to decide decision Anthropic Added a footnote acknowledging Mononoke's influence after Rain's posts gained traction, claim 3
- Who's exposed exposure Rain, former Meta engineer Says she helped design Mononoke and that Scion is a near-complete rebuild of it, claim 8
- What doesn't add up contradiction Rain's initial post Suggested Claude may have independently arrived at the same solution; noted Mononoke is open source, claim 18
- What doesn't add up contradiction Rain's later account Says she received private confirmation the copying was intentional, citing personnel overlap, claim 16
- What it sets up precedent Meta's source control team Built Mononoke from 2017 to 2022, per the write-up, claim 12
| Who | How | Kind | Claim |
|---|---|---|---|
| Anthropic's Scion post | Credited Sapling, Meta's open-source VCS client, for 'hard-won lessons' but never mentioned Mononoke | exposure | 7 |
| Anthropic | Added a footnote acknowledging Mononoke's influence after Rain's posts gained traction | decision | 3 |
| Rain, former Meta engineer | Says she helped design Mononoke and that Scion is a near-complete rebuild of it | exposure | 8 |
| Rain's initial post | Suggested Claude may have independently arrived at the same solution; noted Mononoke is open source | contradiction | 18 |
| Rain's later account | Says she received private confirmation the copying was intentional, citing personnel overlap | contradiction | 16 |
| Meta's source control team | Built Mononoke from 2017 to 2022, per the write-up | precedent | 12 |
What happened
- Anthropic's Scion post credited Meta's open-source Sapling client for 'hard-won lessons' but never mentioned Mononoke, according to the DEV Community write-up.
- Rain says she later received private confirmation that the copying was intentional, citing personnel overlap between the Scion team and people who knew Mononoke's design well.
- Her follow-up post drew 230 points on Hacker News before Anthropic changed its page.
Why it matters
- Who has to decide decision A team weighing Scion's design is also weighing Mononoke's, so it should trace each borrowed concept, such as derived data and the changeset model, to its origin and record that in its own design document.
- What it limits constraint The write-up's author argues an LLM that rebuilds a system emits the design without its sources, so lineage has to be added by the people who publish it.
- What doesn't add up contradiction Rain's first post allowed that Claude might have converged on Mononoke's design alone, while her later account cites private confirmation of intent. The footnote, as described, acknowledges influence, and that wording fits both accounts.
Scion's pitch was scale and speed: a system for 30,000 code merges a day from AI agents, built in five months [1]. Whether the design is Anthropic's is the open question, and the evidence comes in two grades.
The first grade is vocabulary, and it can be checked. Rain says she described Mononoke's "derived data" concept in her 2022 Oxide candidate materials, and that Anthropic's post used the same terminology and framing [9]. Her first post noted that Mononoke is open source [18], so the comparison needs no insider access.
The second grade is intent, and it rests on one private source. Rain says she received private confirmation that the copying was deliberate, citing overlap between the Scion team and people who knew Mononoke's design well [16]. The write-up does not say who confirmed it. Her first post had offered a gentler reading: Anthropic pointed Claude at the problem and Claude arrived at the same solution [18].
Anthropic's side is thin in this record. The footnote exists [3] and Nick G. of Anthropic replied on Hacker News [5], but the write-up quotes neither.
The write-up is also loose with its own dates. It says Meta's source control team built Mononoke from 2017 to 2022 [12], which is five years [15], and elsewhere says the team spent ten years on it [13]. Rain puts the design work in 2018 [8].
The objection, as Rain states it, is about credit. "I am completely fine with good ideas being copied. I am not fine with the serial numbers being accidentally stripped off, and I am certainly not fine with them being intentionally stripped off," she said [2]. She also said, "As LLMs continue to assist in designing new systems, they are likely to index heavily on existing designs" [11].
The write-up's author notes that Scion reverses the syllables of Bonsai, Mononoke's internal changeset model, and allows that it might be coincidence [17]. It would be a tidy one.
After the footnote, Rain's assessment stayed sharp. "This remains yet another shameful example of large AI companies appropriating the commons and presenting others' work as their own," she said [6].
What to watch
- The exact wording of Anthropic's footnote and Nick G.'s Hacker News reply, which would show whether Anthropic credits Mononoke's data model or only its influence.
- Any response from Anthropic or Meta on the claimed personnel overlap between the Scion team and Mononoke's designers.
- Whether Scion's own documentation adds a lineage note naming Mononoke and the Bonsai changeset model.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence25
- Adoption
- Not scored
- Hype gap+40
- Incentives
- Not scored
- Confidence30
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Anthropic published a blog post presenting Scion as its solution to agent-scale version control; Scion was built to handle 30,000 code merges per day from AI agents, in five months.
- [2]
Rain said: 'I am completely fine with good ideas being copied. I am not fine with the serial numbers being accidentally stripped off, and I am certainly not fine with them being intentionally stripped off.'
- [3]
After Rain's posts gained traction, Anthropic added a footnote to its original blog post acknowledging Mononoke's influence.
- [4]
Rain's follow-up post drew 230 points on Hacker News.
- [5]
Nick G. from Anthropic responded on Hacker News.
- [6]
Rain said: 'This remains yet another shameful example of large AI companies appropriating the commons and presenting others' work as their own.'
- [7]
Anthropic's blog post credited Sapling, Meta's open-source VCS client, for 'hard-won lessons' but never mentioned Mononoke.
- [8]
Rain, a former Meta engineer, says she helped design Mononoke, Meta's version control server, in 2018, and that Scion is a near-complete rebuild of it.
- [9]
Rain says she described Mononoke's 'derived data' concept in her 2022 Oxide candidate materials, and that Anthropic's blog used the same terminology, framing and architectural decisions; side-by-side screenshots showed nearly identical diagrams.
- [10]
Rain says the resemblance extends beyond 'derived data' to the entire data model, the Bonsai changeset concept, the server-side API, and multiple architectural decisions considered wildly contrarian when Meta built them in 2018.
- [11]
Rain said: 'As LLMs continue to assist in designing new systems, they are likely to index heavily on existing designs.'
- [12]
The write-up says Mononoke was built by Meta's source control team from 2017 to 2022.
- [13]
The write-up also says Meta's team spent ten years building and productionizing the system.
- [14]
The write-up's author argues that when an LLM rebuilds an existing system it does not cite sources, whereas a human engineer adapting prior art will naturally mention the influence.
- [15]
The write-up's 2017 to 2022 span for Mononoke is five years.
- [16]
Rain says she received private confirmation that the copying was intentional, with personnel overlap between the Scion team and people who knew Mononoke's design intimately.
- [17]
The write-up's author says the name Scion reverses the syllables of Bonsai, Mononoke's internal changeset model, and that this might be a coincidence.
- [18]
Rain's initial post was charitable: it suggested Anthropic may have pointed Claude at the problem and Claude independently arrived at the same solution, and noted that Mononoke is open source.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- dev.toWhen AI Rebuilds Your Work Without Credit: The Scion-Mononoke Story
1 article · October 15, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- AI attribution and creditFollow
- AI Coding AgentsFollow
- Version control systemsFollow