Anthropic's latest robotics experiment reveals a striking shift in AI capabilities. When researchers tested their newest Claude model against the same robotic tasks that previous human-AI teams struggled with, Claude completed everything roughly twenty times faster than the fastest humans. On the four tasks both teams finished, it was more than thirty-seven times quicker than humans working alone. There's a catch: Claude still falters at the delicate touch needed to physically move a beach ball into a target zone—something that requires the kind of intuitive feedback humans learn through practice. But the broader arc is striking. AI first helped humans accomplish tasks. Then humans had to help AI. Now, models are largely solving these problems by themselves. Anthropic researchers describe this as the emergence of physical agentic AI—systems that can manipulate the real world with approaching autonomy.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton