AI programs 38 times faster.

A test that threatens the jobs of today's programmers.

What would happen if the most highly valued and strategic profession of the last decade were quietly obliterated by an artificial intelligence capable of working 38 times faster than elite human teams? An experiment conducted by Anthropic shook the foundations of the tech market, revealing that its new language model—controlling a physical robot—solved complex software engineering tasks in a mere nine minutes.

In the so-called "Fetch" project, the model Claude was granted access to a robot's full development environment, enabling it to analyze data, write code, run tests, identify bugs, and implement fixes automatically. In other words, it functioned as a true engineering agent capable of completing entire development cycles with virtually no human intervention.

The secret to that speed lies in what is known as agentic architecture.

Instead of simply generating a response and waiting for a new instruction, the system operates continuously within a cycle of observation, planning, execution, and validation. At each stage, it verifies whether the code actually worked, identifies potential errors, and automatically adjusts its strategy when necessary. During testing, for instance, Claude was able to locate compatible libraries to control sensors, integrate cameras and LiDAR systems, program computer vision algorithms, and calibrate robot movement—all while autonomously correcting faults that arose along the way.

The most impressive aspect is that this capability was not developed exclusively for Robótic. According to Anthropic, it is an emergent ability of the most advanced artificial intelligence models; as these systems become more capable of reasoning, understanding context, and executing tasks in sequence, they naturally progress to solving increasingly complex problems—a shift that completely transforms the role of software engineers.

Instead of writing every line of code manually, the professional shifts to defining objectives, overseeing results, validating solutions, and guiding artificial intelligence during much larger-scale projects. Does this mean programmers will disappear? Most likely not, but the profession is certainly entering a new phase; just as calculators did not put mathematicians out of work and CAD software did not eliminate engineers, artificial intelligence is poised to profoundly transform the way software development takes place.