Jack Clark, Anthropic’s co-founder, returned from paternity leave in February to find colleagues hardly wrote code anymore, managing five or six copies of the company’s AI, Claude, which sometimes managed several more Claudes. Clark saw this as an early form of recursive self-improvement, where AI accelerates its own development, potentially compressing years of progress into months and leaving society little time to absorb consequences like job disruption or engineered pathogens. He posted blog posts, delivered a talk in England in May, and led an Anthropic report in June titled “When AI Builds Itself,” arguing the technology is already accelerating its development. The report noted the volume of code produced per person at Anthropic has increased eight-fold, with Claude writing 80%. Skeptics, including AI skeptic Gary Marcus, dismissed the claim, saying “all they have really shown is just faster coding,” and noted technological progress has always compounded. Clark conceded coding volume is a crude yardstick, as Claude’s code can be long-winded. Neither Claude nor any large language model is written in code; researchers set growth conditions and pour text through it, letting it adjust itself billions of times. Progress depends on trial and error, and the people building the technology are nearly as much in the dark as everyone else.