Google rolls out Gemini 4 Argon broadly
It scores 77.9% on the DeepSWE software engineering benchmark, currently the top result.
Tip
The “best at coding” ranking changes almost every week; how should ordinary users read these scores?
Anthropic open-sources Scion, a Git replacement built with Claude
A version control system developed with Claude over five months.
Tip
AI can now build a basic tool from scratch; what did the humans actually do?
Meta says hackers stole code and data on 1,500 people via GitHub
The attack came through GitHub and exposed employee information and company code.
Tip
Why code repositories have become an entry point for attacks, and what individuals using GitHub should watch for
AI agent security startup Pluto raises $42M
It focuses on monitoring and protecting AI agents running inside companies.
Tip
As agents multiply, “who watches the agents” is becoming a business
Anthropic and OpenAI want a veto over external evaluator reports
The two companies sought control over whether third-party AI safety assessments get published.
Tip
If the company being evaluated can block the report, is the evaluation still independent?
⚠️ Compiled by an AI assistant from public AI news roundups; original articles are not individually verified. For reference only.