Anthropic's internal code merges have multiplied by ten since May 2025, according to the company. The surge coincides with Claude, Anthropic's AI model, handling more coding tasks directly, and a shift to Scion infrastructure that the company says supports the increased workload.
The change is internal but it signals a broader trend: AI-driven coding is accelerating and reshaping how software gets built. Anthropic, which develops Claude, is now using its own model to write and merge code at a pace that would have been unthinkable a year ago.
What the 10x jump actually means
A merge is what happens when a developer's code changes get approved and added to the main project. Ten times more merges means ten times more code changes landing in Anthropic's repositories. That's not just a productivity metric; it's a shift in how the work gets done.
Claude isn't just suggesting lines of code anymore. It's taking over the keyboard for coding tasks, according to the company. Developers describe a workflow where the model drafts, refactors, and even tests code, leaving humans to review and guide. The result is a faster loop from idea to deployed feature.
Why Anthropic moved to Scion
To handle the increased volume, Anthropic shifted its infrastructure to Scion. The company hasn't detailed the technical specifics, but the move was necessary to keep up with the demand. More merges mean more compute, more storage, and more coordination. Scion provides the backbone for that scale.
The infrastructure change is a practical response to a real bottleneck. When your daily merges jump 10x, your old systems start to creak. Anthropic's decision to migrate suggests the volume wasn't a temporary spike but a new normal.
The ripple effects on software teams
Anthropic's experience is a preview of what's coming to other engineering organizations. If a single company can 10x its merge rate with AI, the competitive pressure on everyone else will be intense. Teams that don't adapt may find themselves outpaced by those that do.
But the shift raises questions about code quality, security, and the role of human developers. Anthropic hasn't shared data on error rates or review times. Without that, it's hard to say whether the 10x jump is sustainable or if it's trading speed for bugs.
Still, the direction is clear. AI coding tools are moving from autocomplete to autonomous contributors. Claude's expanded role at Anthropic is one of the most concrete examples yet.
What to watch next
Anthropic hasn't announced when it will release more details about its internal metrics or the Scion infrastructure. The company also hasn't said whether the 10x merge rate will hold or grow. For now, the number stands as a marker of how fast AI is changing the craft of software development.
Developers outside Anthropic are watching closely. If the model can sustain this pace without sacrificing quality, it won't be long before other companies try to replicate it. The next few months will show whether Anthropic's internal experiment becomes an industry blueprint.
