Skip to content
Original
OpinionSimon WillisonRelayed·· 3 days agoPicks

How Anthropic redesigned source control with Scion

Redesigning source control at Anthropic

Who
Simon Willison
What
discussed Anthropic’s reasons for building Scion and its use of adversarial review
Quote
Interesting piece from Anthropic describing why and how they built their own custom alternative to Git, based on the dramatic increase in rate of changes to their code after the past six months.

What happened

Simon Willison highlights Anthropic’s explanation of Scion, its custom alternative to Git, as merged code changes grew from around 2,500 per day in May 2026 to over 30,000 per day in September. Anthropic reports almost 100 GB of main-branch history and roughly 25 million daily CI checkouts; seven people built the system, completing migration in five months. Willison speculates that Fable/Mythos class models contributed to the increased coding velocity and connects Anthropic’s experience with adversarial review to his own, including the need for engineers to steer models away from over-engineering.

Why it matters

Anthropic’s checkout volume and repository size offer concrete context for understanding how increased coding activity can put pressure on the infrastructure supporting code review and CI.

The story so far

Anthropic built Scion, its own version control system, and fully migrated its monorepo from Git in five months to address infrastructure bottlenecks as daily merged code changes rose from around 2,500 in May to more than 30,000 in September. Anthropic said seven people built the system, with its repository approaching 100 GB of main-branch history and CI performing roughly 25 million checkouts a day. Scion retrieves files on demand, moves operations to the server, and uses layered caches with reported hit rates above 99%. Anthropic reported that median agent workspace startup time fell from 7.1 minutes to 5.1 seconds. In his explanation of Anthropic’s engineering write-up, Baoyu emphasized that Scion is intended for exceptionally large monorepos and that the company does not plan to make it a general-purpose Git replacement.

Keep reading

Latest daily · Issue 24NVIDIA and AI4Bharat Introduce Nemotron-pretraining-Indic-v1 and an Open Data Generation Pipeline

Trending now

  1. 1Meta Releases Muse Spark Open With Mac and Cloud Support, Topping US Open-Weight Index
  2. 2Google Launches Gemini 4 Argon Globally, Expands Paid Access as Benchmark Results Vary
  3. 3OpenAI Updates GPT-6 Astra Pro in ChatGPT With Parallel Agents and Faster Responses
  4. 4Meta’s New Business Terms Raise Dispute Over Website Crawling and AI Training Rights