In a striking reflection on the accelerating pace of artificial intelligence, Andrej Karpathy has described his current state of mind as a kind of “psychosis,” pointing to the disorienting reality of working alongside increasingly capable AI systems. His remarks offer a rare, introspective glimpse into how even leading figures in the field are grappling with the profound changes brought about by the technology they helped build.
Karpathy, a prominent AI researcher and former leader in deep learning initiatives, explained that the turning point came in December, when he realized just how dramatically his workflow had shifted. What once involved carefully writing and refining lines of code had transformed into something almost unrecognizable. He found that he was personally writing only about 20 percent of his code, while the remaining 80 percent was being generated or assisted by AI agents.
This shift, while highly efficient, triggered a deeper sense of unease. For Karpathy, the concern is not simply about productivity gains, but about the changing nature of human involvement in creative and technical work. Coding, long considered a deeply human craft that blends logic, problem-solving, and creativity, is now increasingly mediated by machines that can anticipate, generate, and even optimize solutions with minimal human input.
Describing the experience as a “state of psychosis,” Karpathy clarified that he was using the term metaphorically—to capture the surreal and almost unreal feeling of this transition. The boundaries between human thought and machine output are becoming blurred, creating a sense that the traditional relationship between programmer and program is being fundamentally rewritten.
His comments highlight a broader transformation unfolding across the tech industry. Developers are no longer just builders; they are becoming supervisors of intelligent systems. Instead of writing code line by line, many are now guiding AI models, reviewing their outputs, and iterating on suggestions. This new workflow has dramatically increased speed and efficiency, but it has also raised questions about skill retention and long-term dependency.
One of Karpathy’s central concerns revolves around the idea of agency. When most of the work is done by AI, what does it mean to truly “own” the outcome? If a machine writes the majority of a program, can the human still claim authorship in the traditional sense? These questions are not merely philosophical—they have real implications for how work is valued and how expertise is defined in an AI-driven world.
There is also a cognitive dimension to this shift. As developers rely more heavily on AI tools, there is a risk that their own problem-solving abilities may weaken over time. The convenience of instant solutions can discourage deeper engagement with complex problems, potentially leading to a gradual erosion of foundational skills. Karpathy’s unease reflects this tension between short-term efficiency and long-term capability.
At the same time, he acknowledges the immense power and potential of these tools. AI agents can dramatically reduce the time required to build and deploy software, enabling individuals and small teams to accomplish what once required large organizations. This democratization of technology has the potential to unlock new waves of innovation, lowering barriers and expanding access to advanced capabilities.
However, the speed at which these changes are occurring adds another layer of complexity. The rapid evolution of AI systems means that individuals and institutions have little time to adapt. What feels cutting-edge today may become obsolete tomorrow, creating a constant sense of instability. For someone like Karpathy, who has been deeply involved in the development of AI, this pace can feel both exhilarating and overwhelming.
His reflections also touch on a broader societal concern: the future of work. As AI systems take on more complex tasks, the role of humans in many industries is being redefined. While some jobs may be augmented or enhanced, others could be displaced entirely. The challenge lies in navigating this transition in a way that preserves human dignity and purpose.
Importantly, Karpathy’s perspective is not one of outright pessimism. Rather, it is a call for awareness and reflection. The “psychosis” he describes is not a rejection of AI, but an acknowledgment of how deeply it is reshaping human experience. It is a reminder that technological progress is not just about what machines can do, but about how those capabilities affect the people who use them.
As AI continues to integrate into everyday workflows, the questions raised by Karpathy are likely to become more pressing. How do we maintain a sense of control and understanding in a world where machines handle much of the complexity? How do we ensure that human creativity and critical thinking remain central, even as automation increases?
For now, there are no easy answers. But voices like Andrej Karpathy are helping to frame the conversation in a way that goes beyond hype and fear. By articulating the emotional and intellectual challenges of this moment, he is shedding light on what it truly means to live and work in the age of artificial intelligence.
As the line between human and machine continues to blur, one thing is clear: the future of technology will not just be defined by innovation, but by how we choose to understand and navigate its impact.
