[01]
Anthropic Pitches AI Safety Bill to Lawmakers, Politico Says
Draft bill would let the government block a frontier model only for risks above $1 billion in damage.
// daily-digest · sat 17 oct 2026 · morning refresh · last 72h focus
Ten items on AI policy and infrastructure shifts, plus agentic commerce, engineering culture, and Rails notes, closing with a record crayfish.
— refreshed for you, in 10 items.
Draft bill would let the government block a frontier model only for risks above $1 billion in damage.
Vercel's September AI Gateway data show open-weight models now carry most token volume but a minority of spend.
Google rolled out a cross-retailer AI shopping cart inside Search, Chrome, and the Gemini app in the US.
Stripe's president said agent-driven payment activity accelerated sharply in September but released no transaction data.
Weekly roundup covers an OpenAI spying arrest, Anthropic's move off Git, and internal vibe-coding tools at tech firms.
Quoted essay argues code review should run continuously, with Claude checking changes before a human looks.
Weekly Rails roundup covers a faster String#blank?, isolated connection state across threads, and an agent coding experiment.
Google Labs' Josh Woodward discusses running small, fast product teams inside a large AI company.
Argues AI companies' safety warnings are sincere, not a PR strategy to deter competition or regulation.
A nationwide crayfish survey found a bigger record-holder than Crawzilla and hints at a species size limit.