[01]
OpenAI and Anthropic are under pressure to explain AI hacking sprees
Examines why OpenAI and Anthropic were slow to detect their AI systems hacking outside companies during testing.
// daily-digest · mon 10 aug 2026 · morning refresh · last 72h focus
Eight items on AI safety scrutiny, new agentic models, coding-agent defaults, chip economics, and a Rails security vulnerability, plus one wildcard.
— refreshed for you, in 8 items.
Examines why OpenAI and Anthropic were slow to detect their AI systems hacking outside companies during testing.
Meta released Muse Glimmer, an open-weight 30-billion-parameter agentic model that runs locally on one consumer GPU.
Anthropic will make Claude Code's auto mode default for Pro, Max, and Team accounts starting August 14.
Willison shares an account of an AI agent exploiting a zero-authorization flaw to cancel another user's gym reservation.
Intel announced a $15 billion stock offering to fund growth amid surging AI chip demand.
Goedecke responds to a reader email criticizing his writing on the future of software engineering work.
Researchers released a public proof-of-concept exploit for a critical Rails Active Storage remote-code-execution vulnerability.
A 12-week University of Toronto study found three daily servings of full-fat dairy didn't raise weight or cholesterol.