Anthropic says seven engineers built Scion, a replacement for Git in the company’s monorepo, with Claude over five months. The project comes as Anthropic’s daily merges rose from 2,500 in May to more than 30,000, making source control a development bottleneck.
Anthropic Watch analysis
What happened
Anthropic’s engineering post describes Scion as a response to the strain its growing merge volume put on source control. The company says daily merges increased more than tenfold, from 2,500 in May to over 30,000, and that seven engineers built Scion with Claude in five months.
The announcement makes this more than a tale of an AI helping write code: it connects Claude-assisted development to a piece of infrastructure Anthropic says it needed as its own software work scaled. Read Anthropic’s engineering post.
Why it matters
The figures offer a concrete example of the pressures that can follow faster, higher-volume software development. If a team can produce and merge changes at a much greater rate, the systems that store, coordinate and review those changes may need to keep pace too. Anthropic says source control had become its biggest development bottleneck.
The post also gives a real-world account of Claude being used in building a substantial internal tool, rather than offering another general claim about AI coding. It does not, on the evidence here, tell us how much of the work Claude did or how Scion compares with Git in use. Those are important distinctions between a striking project description and a demonstrated advantage.
Our read
Scion is interesting because the problem and the reported scale are specific: more than 30,000 daily merges, with source control identified as the choke point. That makes this a useful glimpse of how AI-assisted development could change the demands on software infrastructure, as well as the work of building that infrastructure.
The numbers are Anthropic’s account, not a neutral benchmark. The next useful evidence would be how Scion performs in practice and what role Claude played across those five months. Building a Git replacement is no small side quest; the proof is whether it makes the bottleneck smaller without creating a fresh one.
What to watch
- How Anthropic describes Scion’s design and its differences from Git.
- Whether the company shares performance results or operational experience.
- What further detail emerges about Claude’s contribution to the project.
Discussion spark: If AI-assisted coding raises the volume of changes, should software teams first invest in faster source-control infrastructure or in stronger review and testing?
Sources and evidence
Anthropic Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by Anthropic.