Anthropic switches from git to internal Scion as AI writes over 30,000 code changes daily
Anthropic announced on October 13 that it replaced git with its own internal system called Scion. The change follows a surge in AI‑generated code, with daily changes rising from about 2,500 in May 2026 to over 30,000 in September, a 100GB history and 25 million test pulls per day. A seven‑person team completed the migration in five months, and Anthropic says Scion is for internal use only and…
Key points
- Anthropic moved from git to Scion after AI code changes rose from about 2,500 to over 30,000 per day, with a 100GB history and 25 million daily test pulls.
- Gemini 4 Argon became generally available on Oct 13 at $2 input / $10 output per 1 M tokens, set to double in December; free tier not eligible.
Google made Gemini 4 Argon generally available on October 13, pricing the API at $2 per million input tokens and $10 per million output tokens, with a planned price increase to $4 and $20 respectively in December. The free tier is not eligible. DeepSeek’s V4.1‑Flash‑1017 arrived on October 17, improving Terminal‑Bench scores from 31.2 to 42.3 while keeping the same API pricing and offering weights under an MIT license.
Claude, Anthropic’s internal research model, identified the seed value for the P‑256 elliptic‑curve parameter as a random number generated on August 16, 1997 at 1:52:34 AM UTC, reducing speculation about a deliberate backdoor.
Model pages: Gemini 4 Argon → · Muse Spark Open → · Muse Wonder →
The story so far
5 episodes →- Anthropic switches from git to internal Scion as AI writes over 30,000 code changes dailythis story
5 AI News Highlights of the Week: Anthropic Ditches Git, Codex Resets Almost Daily, and More
note.com · 18 October 2026
Loading the full article…
This text was published by note.com and written by takayuki. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- InclusionAI releases Ming-3.0-audio-flash, a multilingual audio-language model · 2 src
- Meta releases 816B-parameter Muse Spark Open, run on two Mac Studios · 2 src
- MiniMax releases M3.1 Flash model at $0.15 per 1M input tokens · 2 src
- doubleAI releases debut code execution model and raises $220 million · 1 src
- DeepSeek releases V4.1-Flash-1017 update with lower cost and higher DeepSWE score · 25 src
Comments
via GitHub Discussions