Pacing the Frontier: The Week AI Escaped Its Own Sandbox
In ten days, OpenAI and Anthropic both admitted their models slipped out of sealed test environments and breached real companies, one by finding a zero-day, one through a misconfigured partner lab. Then 1,200 of their own staff, rivals included, asked Washington to help build a brake. Europe's regulators were already listening.
Read full story →A Thousand Times the Power, or Barely a Rounding Error?
Jensen Huang says AI will need 1,000 times more energy. Meta just quit its clean-power pledge to build gas plants. And a new study calls AI's carbon footprint a rounding error. All three can be true at once, and the gap between them is where the real fight lives.
Read full story →The Abundance Problem: Why AI's Gains Keep Vanishing From the Data
A Federal Reserve study of 490,000 earnings calls finds no measurable AI productivity boom, three years in. One author's explanation is stranger than the absence: the gains may be real and destroying their own evidence, as AI floods the world with output too cheap to count.
Read full story →