Anthropic has announced the second phase of Project Fetch, an experiment that began in August 2025 to explore how its AI model, Claude, could control off-the-shelf robotic quadrupeds (robodogs) to perform sophisticated tasks. The initial phase involved Anthropic employees without robotics expertise using Claude to direct the robots in amusing and complex activities.
Phase two builds on these foundations, with verified reports indicating that the project now focuses on more advanced capabilities, such as autonomous navigation and object manipulation. The experiments aim to demonstrate how AI can bridge the gap between human intent and robotic action, potentially enabling non-experts to deploy robots for practical applications.
According to Anthropic's official communications, the project has shown promising results in task completion rates, though specific metrics have not been disclosed. The company emphasizes that these are research experiments and not yet commercial products.
The development highlights the growing trend of integrating large language models with physical robotics, a field that has seen increased investment from tech companies. Anthropic's approach differs by focusing on accessibility, allowing users with minimal training to command complex robotic systems.
