Anthropic's AI agent outperformed engineers by 20 times in controlling a robot dog — a new frontier for physical AI
The artificial intelligence market continues to surprise: Anthropic has presented the results of the second phase of its Project Fetch experiment. The Claude Opus 4.7 model demonstrated phenomenal performance in configuring and controlling a four-legged robot, completing tasks 20 times faster than teams of human engineers. This is not just a number — it is a qualitative shift in understanding how neural networks can interact with the physical world.
Let me remind you that in August 2024, Anthropic employees with no experience in robotics tried to program a robot dog using AI. At that time, the model only acted as an assistant, speeding up the search for solutions. Today's version of Claude Opus 4.7 operates almost completely autonomously under minimal researcher supervision. The neural network independently completed a full cycle of tasks:
- connected to video sensors and LiDAR;
- wrote a program for manual control;
- created a robot path monitoring system;
- configured an object recognition algorithm.
The results are impressive: Opus 4.7 turned out to be 18 times faster than a team using previous AI versions, and 37 times faster than people working without chatbot assistance. Moreover, the neural network generated code that is 10 times smaller in volume than that of human teams. This speaks not only to speed but also to the quality of solutions — the efficiency of Claude's algorithms is significantly higher.
The key point: Anthropic did not introduce specialized algorithms for controlling hardware. The progress in robotics was a side effect of the general scaling of language models. This confirms my long-standing hypothesis: fundamental improvements in AI training directly translate into new areas of application, including physical interaction.
However, one should not think that AI has already completely replaced humans. Claude still struggles with precise physical actions. During the experiment, the model managed to guide the robot to the target but failed to carefully push a ball to the exact spot. This requires complex real-time feedback — an area where humans still maintain superiority.
Anthropic believes the industry is entering an era of "physical AI agents." My analysis shows: this is not just a marketing term. Already, neural networks can use standard tools and equipment as effectively as software code. The only question is how quickly the remaining limitations will disappear.
Expert comment: Anthropic's breakthrough is a signal for the entire market. If the current pace of development continues, in the next 2-3 years we will see the first commercial products where AI manages physical processes on par with humans. Investors should pay attention to startups in robotics and industrial automation — this is where the next wave of growth will be concentrated.