At the Sequoia AI 2026 Conference, Andrej Karpathy highlighted a significant shift in the role of AI. He noted that manual modification of AI outputs has become almost obsolete. While many see AI as a tool for efficiency, the conference emphasized a deeper transformation: the rewriting of operational rules. The way tasks are executed is changing; instead of detailing every step, one now simply sets a goal, and AI models complete the task autonomously. This evolution challenges traditional software engineering, which relied on breaking down tasks into explicit steps. As AI models grow more capable, the need for intermediary applications diminishes, redefining the value of tools focused on format conversion and low-code platforms.
The division of labor between humans and machines is also evolving. Intelligent agents can handle many tasks autonomously, though they still require human oversight for complex decision-making. The key determinant of what tasks machines will take over is verifiability. Tasks with clear standards, like code running or data formatting, are easily automated. However, tasks requiring nuanced judgment remain human-centric. This shift affects various professions, from design to law, where automation is taking over routine tasks, leaving humans to handle strategic decisions.
Looking ahead, Demis Hassabis predicts that Artificial General Intelligence (AGI) will be realized by 2030. Entrepreneurs are advised to focus on areas like materials science and medicine, where deep understanding of the physical world offers a competitive edge. As AI capabilities expand, the human role shifts from executing tasks to understanding their significance. The real value lies in defining goals and understanding industry needs, a role AI cannot fulfill. As we approach AGI, the ability to discern what is truly important will become increasingly vital.
