Anthropic Replaces Git with Custom Version Control System, Signaling AI-Driven Code Velocity Challenges

Anthropic, a leading AI research company, has announced a significant shift in its internal development infrastructure: the complete migration of its monorepo from Git to a new, purpose-built version control system named Scion. This change was necessitated by an unprecedented surge in code merges, primarily driven by the company's Claude-powered coding agents. Anthropic reported an increase from approximately 2,500 daily code merges in May 2026 to over 30,000 by September 2026, a more than tenfold jump in just four months. This rapid acceleration rendered Git a significant bottleneck in their development workflow, prompting the creation of Scion. This development is highly significant for practitioners in cloud, DevOps, and AI. It underscores a critical emerging challenge: the infrastructure supporting software development is struggling to keep pace with the velocity of AI-generated code. While AI coding tools promise increased productivity, their output can overwhelm existing version control and CI/CD pipelines. For DevOps teams, this means that optimizing code generation is only half the battle; the other half is ensuring the underlying systems can absorb, manage, and deploy that code efficiently. Organizations heavily investing in AI-driven development, particularly those with large monorepos or complex codebases, will likely encounter similar scalability issues with their current source control solutions. This move by Anthropic fits into a broader, well-established trend in cloud and DevOps where infrastructure itself becomes a critical enabler (or constraint) for innovation. Just as Infrastructure as Code (IaC) became essential for managing cloud resources at scale, and GitOps emerged to apply Git-based workflows to infrastructure, the rise of AI-driven code generation demands a re-evaluation of fundamental development tooling. The IaC market itself has seen significant growth, exceeding $1.4 billion in 2026, with a 24% CAGR, indicating a widespread reliance on code-driven infrastructure management. However, the focus has largely been on the *creation* of infrastructure code. Anthropic's experience highlights that the *management* of rapidly evolving and high-volume code, regardless of its origin, is now a frontier for innovation. The industry has been grappling with the implications of AI in IaC workflows, with discussions around AI's role in code generation and drift detection. Anthropic's announcement shifts the conversation to the foundational layer of code management itself. In practice, this means that organizations leveraging AI coding agents should proactively assess the scalability and performance of their existing version control systems. Practitioners should monitor their merge rates and repository sizes, and consider whether their current tooling can sustain a tenfold increase in code changes. While Anthropic has stated that Scion will not be publicly released or open-sourced, its existence signals a demand for more robust, high-throughput version control solutions. This could lead to new commercial offerings, open-source projects, or specialized tooling designed to handle the unique demands of AI-generated code. Furthermore, security teams should note Anthropic's decision to build a dedicated IAM layer for Scion, rather than relying on Git's existing access controls, indicating a growing need for purpose-built guardrails for agent-to-repository access in the AI era. The trade-off between the flexibility and ubiquity of Git and the performance requirements of AI-driven development will be a key consideration for many organizations moving forward.