The AI industry's grand vision of self-improving machines may hit a roadblock. A new study reveals that AI agents struggle with open-ended research, the kind that requires creativity and judgment for genuine breakthroughs.

Without this capability, it’s unclear if AI can truly improve itself without human intervention. The big question now is: will AI revolutionize technology or just repeat its own errors ad infinitum?

In other tech news, OpenAI has paused some model work due to safety concerns, setting it apart from Anthropic's approach. Meanwhile, China's Unitree has made a splash in the stock market, becoming a key player in humanoid robots and potentially fueling the Sino-US tech war.

Elsewhere, Meta faces accusations of deliberately hooking children on its platforms, while smart glasses for law enforcement raise ethical red flags. X’s algorithm learns your dislikes to show you more of it, and even ICE is reconsidering its stance on smart glasses due to privacy concerns.