Anthropic launches a free OSS vulnerability scanner and a critical-infrastructure defense program, while Google turns Gemini into a persistent agent that can run Claude underneath. Plus OpenAI disrupts a Category 5 Russian influence operation, Sierra and Meta publish an open protocol for personal AI agents, Mistral ships a trillion-parameter model, Rails tightens S3 upload permissions, Simon Willison builds a feature by talking to Codex's voice mode, Scott Alexander rebuts Steven Pinker on AI risk, Noah Smith wonders if the Rapture already happened, and a wildcard: astronomers find a planet reborn from its dead star.

Anthropic · Oct 8

Anthropic launched a free open-source vulnerability scanner and a Claude-staffed defense program for critical-infrastructure providers.

Google · Oct 8

Google Cloud unveiled a persistent Gemini agent for cross-app work tasks that can also run Claude models underneath.

OpenAI · Oct 8

OpenAI disrupted a Category 5 Russian influence operation and an Iranian one using fake journalist personas.

Bret Taylor & Clay Bavor, Sierra · Oct 6

Sierra and Meta published an open OAuth-based standard for how personal AI agents authenticate with businesses.

Anna Heim, TechCrunch · Oct 6

Mistral released Large 4, a trillion-parameter model it positions as a third way between closed and open rivals.

This Week in Rails · Oct 9

Weekly Rails roundup covering new S3 ACL handling, a Postgres maintenance-database option, and frozen ActiveSupport events.

Simon Willison · Oct 9

Simon Willison built a newsletter-archive feature for his blog mostly by talking to Codex's voice mode.

Scott Alexander, Astral Codex Ten · Oct 6

Scott Alexander rebuts Steven Pinker's public arguments dismissing AI safety concerns, point by point.

Noah Smith, Noahpinion · Oct 9

Noah Smith argues algorithm-driven feeds have fragmented mass culture into isolated, personalized experiences.