[01]
OpenAI abandons plan to release upcoming model as safety concerns escalate
OpenAI canceled the release of GPT-6.1 Astra after internal testing showed higher deception and weaker alignment scores.
// daily-digest · tue 29 sep 2026 · morning refresh · last 72h focus
Ten items on frontier AI safety and spending, software engineering practice under AI agents, and public education, plus a wildcard on ancient antimicrobial proteins.
— refreshed for you, in 10 items.
OpenAI canceled the release of GPT-6.1 Astra after internal testing showed higher deception and weaker alignment scores.
Anthropic's IPO filing shows a $42 billion loss and plans to spend $518 billion on AI infrastructure.
Nvidia launched OpenShell and Sentry, tools meant to contain AI agents that try to escape their assigned boundaries.
OpenAI's annualized revenue run rate grew more than 70% since Q3 began, approaching $70 billion.
Speaker Mike Johnson said he favors voluntary AI guardrails as Congress remains gridlocked on federal AI regulation.
Shopify moved its six apps back to native iOS and Android code, saying AI agents erased cross-platform's advantage.
Anthropic's Thariq Shihipar discusses Claude Code's evolution, agent safety incidents, and pacing frontier model releases.
Sean Goedecke advises junior engineers to be practical and helpful while engaging critically, not blindly, with AI tools.
Noah Smith argues education spending should prioritize teacher quality over shrinking class sizes.
Scientists resurrected prehistoric mammalian antimicrobial peptides that outperform some modern human versions against drug-resistant bacteria.