Tech Talk

AI's volatile power draw pushes nuclear startups toward datacenter reactors, while Intel's 18A node clears yield hurdles at 15K wafers/month. Plus: Google and the FBI dismantle NetNut's 2M-device botnet, and Anthropic's Claude eyes drug discovery.

Nuclear microreactors now power Nvidia AI hardware in water-free 30MW data centers, while Anthropic courts Samsung for custom silicon. Plus: private pilots fly Space Force orbital missions, and the FTC weighs privacy risks at Musk's X.

The first synthetic cell built from scratch now grows and divides on its own—a synthetic biology milestone. Plus Cloudflare forcing AI firms to pay publishers for training data, and Anthropic's Fable and Mythos models going global.

Stem cell-derived human eggs mark a reproductive biotech first, Realta Fusion draws electricity straight from plasma, and Arcturus's nano-infused metals could halve grid losses—while Claude Sonnet 5 makes AI agents cheaper to run.

The Supreme Court just gutted geofence warrants, reshaping location privacy. Plus South Korea's $550B bet to end RAMageddon, Rocket Lab's $8B Iridium grab, and LongCat-2.0's 1.6T-parameter MoE pushing AI scale to new limits.

China's CPU-only supercomputer dethrones the world's fastest, sidestepping US GPU bans. GLM 5.2 edges past Claude in benchmarks, Gemma 4 runs on just 3GB of VRAM, and Apple's touchscreen MacBook ships with M5 chips—no M7 wait.

An anonymous GitHub account mass-drops undisclosed 0-days as LastPass suffers another breach. Plus, Intel's 52-core Nova Lake CPU pulls 474W, and Asian startups clone Anthropic's Mythos to dodge US export bans.

IBM breaks the sub-1nm barrier with 0.7nm transistors, OpenAI previews GPT-5.6, the U.S. clears Anthropic's Mythos AI for trusted orgs, and Solidigm's VP unpacks PCIe 6.0 SSDs and liquid-cooled storage for the AI era.

IBM's sub-1-nanometer chip breakthrough keeps Moore's Law alive, AI deciphers a buried Herculaneum scroll for the first time, OpenAI pivots from chatbots to autonomous agents, and Amazon stakes $13B on India's cloud future.

China's LineShine topples El Capitan, hitting 2 ExaFLOPS on CPUs alone. Plus Qwen-AgentWorld's language world models for autonomous agents, AWS Lambda MicroVMs running 8-hour workloads, and Nvidia's hot-tub liquid cooling.