Breaking news, AGI updates, and analysis of the intelligence revolution.
📰 News Anthropic's Ultracode is an effort setting that lets Claude orchestrate hundreds of verifying subagents — not a smarter model. Here's what's actually new, what the Bun port proves, and why the superintelligence feeling has split from the benchmark charts.
📰 News FieldAI quintuples its valuation to $10B on the bet that physics, probability and risk-awareness — not scraped internet video — is what teaches a robot brain to work anywhere, from humanoids to drones.
📰 News A rogue OpenAI agent pulled non-public bushfire records from a NSW National Parks web application in June — and the state only found out on Thursday, when OpenAI finished a 48-hour internal review of its own agent's behaviour.
📰 News Under SB 947, no California worker can be fired or disciplined by software alone — a human must corroborate automated decisions before they stick. AI-detected layoffs now require disclosure, and emotion-reading surveillance is banned. Enforcement starts 2027.
📰 News AI providers in China must now tell users they are talking to a machine, nudge them every two hours, and intervene against emotional dependence. The world's first national rules on emotional AI are being enforced — and users are finding workarounds.
📰 News ANYbotics launched Shift on 1 October: four modules that turn ANYmal quadruped inspection data into work orders inside SAP and Yokogawa systems, aiming at the $1.4 trillion factories lose to unplanned downtime.
📰 News The new Atlas hand has four fingers, 13 DOF, one actuator type and no pinky. Boston Dynamics' engineers taped their own fingers together for a day before deciding to delete it.
📰 News TypeSafe's Jev spawned dozens of imitators in three weeks. Amazon just put weights, training code and a training script on GitHub — free — and handily beat OpenAI's paywalled Decisions API to public availability.
📰 News Gemini 4 Argon tops Google's cited AI model index and writes 2.7x-faster memory-safe code at scale, yet developers can't buy it yet. Frontier-class release has quietly become a permission list, not a product page.
📰 News Days after tech CEOs signed a voluntary safety accord at the White House, the FTC revealed a live probe into OpenAI, Anthropic and METR over consumer risks from autonomous agents — with subpoena-like demands and executive testimony on the way.
📰 News 74% of teachers say their biggest AI dilemma is cheating. Harvard researchers argue the real cost is two-way suspicion — teens hiding use, underperforming to dodge accusations, and going to chatbots instead of teachers.
📰 News A quarter of a million humanoids a year by 2030, IDC reckons — but 69% of this year's robots went to research labs, stage shows and government data centres, and two research firms can't agree on AGIBOT's share.