Anthropic rebuilds source control for AI-scale code

Summary

Anthropic says Claude-driven development pushed daily merges from about 2,500 in May to over 30,000 in September. Human-scale git infrastructure became the bottleneck, so the team built Scion, an in-house version control system, and fully migrated the monorepo within five months. Agents and CI jobs now fetch only the data they need from the server, with new code search and review surfaces built alongside. The merge figures are Anthropic's own, and the post was published October 13 and expanded October 15.

Why it matters for our work

A concrete case where review, builds, and version control—not code writing—become the new bottleneck as AI coding scales. Teams accelerating research and development with AI can expect similar pressure points soon.

Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.