AI companies are nearing “recursive self-improvement,” or RSI, in which AI models improve themselves and build their successor, Fortune reports. Anthropic said this week that its model Claude is leading 26% of its model research and development, completing most of a given task “end-to-end from a high-level prompt” under human supervision. The models are not working completely autonomously yet. Definitions of RSI vary: some companies count any AI feedback on model improvement, while others require AI working toward that goal fully autonomously. Anthony Aguirre, president and CEO of the nonprofit Future of Life Institute and a physics professor at the University of California, Santa Cruz, said autonomous RSI means AI designing the next version of the system, then the next, and so on, and that it gets faster as AI does more of it because AI operates much more quickly than humans. John Thickstun, an assistant professor of computer science at Cornell University, said fear around RSI is based largely on a runaway superintelligence emerging from that process, but a more grounded view suggests recursive self-improvement has been going on for a while, with past models used to write code for systems that create the next generation.