Pulse Daily — Wednesday, August 19, 2026

On Wednesday, August 19, 2026, the AI world reckoned with the fallout of OpenAI's rogue agent breach while celebrating China's Unitree robotics IPO surge, grappling with semiconductor price hikes, Europe's data-center energy crisis, a new ChatGPT teen safety mode, a GitHub Copilot-introduced shell-injection flaw, an AI manager that fired a human worker, Z.ai's open-weight cybersecurity model outpacing commercial rivals, the U.S.-China H200 chip standoff, and the dawning recognition that electricity—not algorithms—has become AI's defining bottleneck. Stories: OpenAI Puts a Speed Limit on Its Most Dangerous Models · Z.ai Ships GLM-5.3 — Open-Weight Model Tops Closed Frontier Models on Cyber · OpenAI Rolls Out ChatGPT for Teens — Age-Prediction Routes Under-18s Automatically · Wiz's AI Agent Caught What GitHub's AI Scanner Missed — Then GitHub Didn't · Unitree Surges 460% on Debut — First Humanoid Maker to List in Mainland China · Memory Chip Shortages Are Pushing AI Build-Outs Beyond US Borders · Europe's AI Bottleneck Isn't Chips or Talent — It's Electricity · An AI Manager Just Recommended Firing a Human — After a Human Had to Remind It of the Rules.