According to a follow‑up post on sunshowers.io, the author — who previously worked on Meta’s Mononoke source‑control system — says that Anthropic’s Scion system is not an independent invention but a deliberate copy of Mononoke’s Bonsai data model. The writer explains that while an earlier article tried to be charitable and suggested the resemblance might be unintentional, private communications after publication confirmed that the similarity was intentional and that there was overlap in personnel between the Mononoke team and those who built Scion.
The post details several technical parallels. It notes that Anthropic describes Scion changesets in a way that matches Mononoke’s Bonsai changesets exactly, and points out that the name “Scion” reverses the syllables of “Bonsai” and drops the initial “B”, a similarity the author characterizes as either a coincidence or a deliberate nod. Beyond the changeset concept, the author says the resemblance extends to the entire data model and to various structures built around it, including indexes for blame and other derived‑data features that are a natural consequence of the Bonsai model.
Regarding the server‑side API, the post states that Mononoke implemented a set of version‑control queries and mutations that other systems had before, but that if one pointed a language model at Mononoke and asked it to rebuild the same functionality, the result would look very similar to what Scion provides. The author emphasizes that the Bonsai model is a write‑ahead log‑style Merkle DAG, a design that ran contrary to the prevailing Git‑like approach when it was created in 2018 and required solving difficult problems such as server‑side rebases. The writer recalls having to defend the novel design to multiple layers of management at Meta and considers the resulting system one of their proudest career achievements.
Because the Bonsai model has remained in use for almost a decade and has been adapted far beyond its original scope, the author argues that anyone familiar with Mononoke would recognize its influence. Given the confirmed personnel overlap, the writer concludes there is no reasonable way the Scion team could have been unaware of Mononoke’s work, and calls the lack of proper attribution a “shameful” example of large AI companies presenting others’ work as their own. The post ends by urging Anthropic to acknowledge the source of its inspiration, noting that while building a system in five months is still an achievement, presenting it as original diminishes the effort of the Mononoke team that spent years developing and productionizing the technology.