What is agentic engineering according to Andrej Karpathy?

Agentic engineering is a new phase of AI software development where AI agents autonomously generate and refine code, with humans acting as orchestrators and supervisors.

How does agentic engineering differ from vibe-coding?

Unlike vibe-coding, where humans directly prompt AI, agentic engineering involves AI agents working autonomously, guided and overseen by human expertise.

When is agentic engineering expected to become prominent?

Andrej Karpathy predicts significant improvements in AI models and agent layers by 2026, making agentic engineering a central part of software development.

Home / Technology / AI Learns to Engineer: Karpathy Unveils Agentic Future

AI Learns to Engineer: Karpathy Unveils Agentic Future

9 Feb

•

Summary

AI agents will autonomously generate and refine code.
Agentic engineering emphasizes expertise in guiding AI agents.
2026 is predicted to see major advancements in AI agent capabilities.

AI Learns to Engineer: Karpathy Unveils Agentic Future

Andrej Karpathy, a founding member of OpenAI and former director of AI at Tesla, has coined the term "agentic engineering" to describe the upcoming era of AI-driven software development. This new paradigm shifts from "vibe-coding," where humans directly prompt AI, to a system where AI agents autonomously write and refine code.

Agentic engineering emphasizes the crucial role of human expertise in orchestrating and overseeing these AI agents. Karpathy highlighted that this process involves an "art & science and expertise" that can be learned and mastered.

Karpathy forecasts that by 2026, advancements in both AI model capabilities and agent technology will make agentic engineering a central component of software creation. This evolution promises increased leverage from AI agents without compromising software quality.