Andrej Karpathy: From Vibe Coding to Agentic Engineering

The transition to Software 3.0 marks a fundamental shift from explicit coding and training neural networks to prompt-based computation, where LLMs function as the primary interpreter. Andrej Karpathy, a pioneer in AI development, highlights that this paradigm enables the automation of complex information processing, extending far beyond simple code generation. While agentic tools significantly boost productivity, they exhibit "jagged" intelligence, excelling in verifiable domains like mathematics and programming while occasionally failing at basic common-sense tasks. Consequently, the role of the engineer is evolving into "agentic engineering," which prioritizes high-level design, aesthetic judgment, and rigorous oversight to ensure quality and security. As intelligence becomes cheap and ubiquitous, the most valuable human skill remains deep understanding, as humans must continue to act as the directors of these powerful, stochastic, and increasingly autonomous systems.