OpenAI Lands a Heavyweight
The AI talent war just claimed one of its biggest names. Noam Shazeer — co-leader of Google's Gemini models and a vice president of engineering — is leaving for OpenAI, a move he announced Wednesday, according to CNBC. Shazeer is among the most prominent figures in the field, and his defection is a striking coup for OpenAI in the escalating contest for elite researchers. It also lands at an awkward moment for Google, which is leaning hard on Gemini across its product line.
Gemini Comes Home
Case in point: Google will release a fully redesigned smart speaker on June 29 for $99.99. The new Google Home Speaker swaps the company's decade-old assistant for a chattier, generative-AI helper built around "Gemini for Home." It's a concrete sign of how quickly the large-language-model boom is reshaping the everyday devices on your kitchen counter.
Robots That Teach Themselves
Nvidia, meanwhile, is pushing automation up the stack. According to Ars Technica, Decrypt, and Tech Times, the company has built a system in which robots help train themselves — with teams of AI coding agents, rather than human researchers, directing the work. The approach hints at a future where the tedious, hands-on labor of robot training is handed off to software. That vision is also drawing capital: Inc.'s "7 Robotics Startups to Watch" report says AI is fueling a robotics startup boom, and investors are following the money.
Not every robot story is rosy. Waymo is recalling more than 3,800 self-driving taxis to patch a software flaw that could steer them into closed freeway construction zones, per Engadget and Tech Times — a reminder that autonomy still has sharp edges.
Wall Street Pumps the Brakes on Claude
Two banking giants are tightening the reins on AI. JPMorgan has stopped its Hong Kong staff from using Anthropic's Claude models, according to the Financial Times, and crypto.news reports Goldman Sachs is moving to limit access for its Hong Kong employees as well. The restrictions show big institutions still wrestling with how — and where — to let powerful AI tools into the workplace.
Washington's Quiet Hand
Policy is murkier than advertised. The Trump administration entered office promising to stay out of the AI industry's way, but Axios reports the reality is more interventionist than the hands-off rhetoric suggests — deregulation in name, with a heavier touch in practice.
The Quantum Race Accelerates
The day's biggest cluster of news came from quantum computing. Amazon Web Services set a 2028 target to deploy its first error-corrected, fault-tolerant quantum computer to the cloud, working through an expanded partnership with QuEra. Across the Atlantic, France made two notable moves: it agreed to buy a quantum computer from homegrown startup Alice & Bob — described as a world first for the underlying "cat-qubit" technology — even as its broader, NVIDIA-powered national AI plan, unveiled a year ago, starts taking physical shape. And Finland's IQM installed its first US quantum computer at Oak Ridge National Laboratory, parking the machine next door to the world's fastest supercomputer.
Taken together, today's stories sketch an industry sprinting on every front at once — chasing talent, automation, and the next computing paradigm — while regulators and risk managers scramble to keep pace.