The AI agent Claude Opus 4.7 outperformed humans by 20 times in controlling a robot dog — a new stage in the evolution of autonomous systems
A key moment in the development of artificial intelligence: Anthropic's Claude Opus 4.7 model demonstrated the ability to independently program and control a four-legged robot, completing tasks 20 times faster than the best teams of human engineers. This is the result of the second phase of the Project Fetch experiment, which fundamentally changes the understanding of AI's capabilities in the physical world.
In August 2024, company employees with no robotics experience attempted to program a robot dog using AI, which only accelerated the search for solutions. Today, the situation is fundamentally different: Claude Opus 4.7 worked almost autonomously, under minimal researcher supervision. The neural network independently completed the full setup cycle:
- connected to video sensors and lidar;
- wrote a program for manual control;
- created a robot path monitoring system;
- configured an object recognition algorithm.
The numbers are impressive: the Opus 4.7 model proved to be 18 times faster than a team using older AI versions, and 37 times faster than humans working without chatbot assistance. Moreover, the neural network wrote more efficient code — its volume turned out to be 10 times smaller than that of human teams. This is direct evidence that modern language models are capable not only of generating code but also of optimizing it at a level inaccessible to humans.
It is important to emphasize: Anthropic did not implement specialized algorithms for controlling hardware. The progress in robotics was a side effect of the general scaling of language models. This means we are observing not a narrow specialization, but a fundamental expansion of AI capabilities.
However, there were limitations. Despite the success, Claude still struggles with precise physical actions. The model managed to guide the robot to the target, but failed at the task of gently pushing a ball to a specific point. This requires complex real-time feedback, in which humans still outperform AI. Anthropic believes that the industry is entering an era of "physical AI agents," where neural networks will be able to use standard tools and equipment as effectively as they currently work with software code.
My analysis: This experiment is not just a demonstration of speed, but a signal of a paradigm shift. If earlier AI was an assistant, now it is becoming an independent operator of complex physical systems. The only question is how quickly we can solve the problem of precise physical feedback — and who will be the first to implement such systems in real-world industry.