
What happened
Anthropic built Scion, its own source control system to replace Git, completed the migration in five months with a seven-person team, and says the system handles its main branch history of almost 100 GB.
Why it matters
Anthropic's code changes merged per day rose from around 2,500 in May 2026 to over 30,000 by September, suggesting its internal tooling may no longer keep pace with AI-assisted coding velocity.
WHO IT HITSEngineering teams at AI labs and large software organizations may find their existing version control and CI tooling strained by AI-assisted coding volumes, and could weigh building custom alternatives akin to Anthropic's Scion.
Summaries like this, in your inbox every morning.
Anthropic's decision to build Scion came after code changes merged per day rose from around 2,500 in May 2026 to over 30,000 in September. The company said its main branch history alone is almost 100 GB and that a compressed working copy is over 25 GB. It also performs on the order of 25 million code checkouts per day for CI, which it said would amount to over 600 PB of git traffic per day if fetched naively.
Simon Willison, writing on his blog, noted that a chart of those changes showed markers for public model releases in 2026 and guessed that the introduction of Fable/Mythos class models was a key factor in the increased coding velocity. He also described Anthropic's new code review system, called Antique, and said Scion sounded to him like a reinvention of centralized source control along the lines of Subversion.
Anthropic also makes extensive use of adversarial review. Willison noted that while this is incredibly effective, it can be myopic, and skilled engineers needed to steer Claude away from over-engineering parts of the system in response to hypothetical concerns.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Simon Willison added Gemini 4 Argon support to llm-gemini, his Gemini API tool

A team leader says AI-driven development has not reduced his workload

GitHub released Stacked pull requests in public preview on July 30, 2026 and made them generally available on…

The developer reproduced false passes locally with Python 3.14 and pytest 9.1 on Windows Git Bash: `pytest tes…

Wells Fargo CEO Charlie Scharf said the bank uses AI and autonomous agents to write code faster, close previou…

Google made Gemini 4 Argon available worldwide today through its API, the Gemini app, and Google Antigravity
