Svmuu News: Anthropic has released the results of the second phase of “Project Fetch,” evaluating the performance improvements of its latest model in real-world robotic operations. The experiment was conducted in August 2025. Researchers had Anthropic employees—who were not robotics experts—use off-the-shelf quadruped robots to complete a series of complex tasks, comparing performance between “using the Claude model for assistance” and “relying solely on humans and the internet.” The results showed that when operated fully autonomously by the latest model, Claude Opus 4.7, it significantly outperformed the human team in average speed across all completable tasks, with execution speeds at least 10 times faster. Anthropic stated that this advancement was not the result of specialized robotics training, but rather an extension of the capabilities of general-purpose large language models. The company noted that AI is gradually transitioning from “assisting humans in using tools” to the early stages of “directly operating physical tools,” following a path similar to the evolution toward agent-based programming in the field of software engineering.