AI agent took control of a robot dog: Claude Opus 4.7 proved faster than an entire team of engineers
The world of artificial intelligence continues to amaze. As part of the updated Project Fetch experiment, Anthropic demonstrated that a new-generation language model can not just assist a human, but completely replace them in a complex engineering task. This involves controlling and configuring a four-legged robot — a robot dog.
During testing, the Claude Opus 4.7 model operated almost autonomously, under minimal researcher supervision. The result is impressive: the AI completed the task 20 times faster than a team of human engineers working with the previous version of the model. Moreover, the neural network was 18 times faster than a group using older AI versions, and 37 times faster than humans working without chatbot assistance.
How It Worked
The model independently performed the full cycle of robot configuration:
- connected to video sensors and lidar;
- wrote a program for manual control;
- created a path monitoring system;
- configured an object recognition algorithm.
Notably, the code generated by Claude turned out to be 10 times more compact than that of human teams. This speaks not only to speed, but also to the high efficiency of the neural network's work.
Progress Without Specialization
The experiment's authors emphasize: success in robotics was a side effect of the general scaling of language models. Anthropic did not introduce specialized algorithms for controlling the hardware. This discovery confirms the hypothesis that universal AI systems can master new physical skills without additional training.
Weaknesses Still Exist
Despite the impressive results, 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. This requires complex real-time feedback — an area where humans still maintain superiority.
Anthropic is confident: the industry is entering an era of "physical AI agents." In the future, neural networks will use standard tools and equipment as effectively as they currently work with code.
Cryptalist Expert Opinion: This experiment is a clear demonstration that the boundary between "digital" and "physical" AI is rapidly blurring. If today a model has learned to control a robot dog without specialized training, then tomorrow it will be able to control drones, machine tools, and medical manipulators. Investors should closely monitor this direction: companies that first integrate such physical agents into the real economy will gain a colossal competitive advantage.