Interest in artificial intelligence (AI) is shifting from casual experimentation to more sophisticated software development, as leading figures encourage new frameworks for building with AI. Andrej Karpathy, known for his work with OpenAI and Tesla Autopilot, has shared new terminology that reflects this evolution. His earlier phrase “vibe coding” inspired a wave of novice coders, but his latest concept, “agentic engineering,” points to a future where the collaboration between engineers and AI agents is at the core of professional software creation. Investors and major brands like Cursor, Lovable, and Replit are following this trend closely, fueling growth across the sector. The influence of language in shaping technological movements and markets is once again evident, with startups adopting these ideas to compete against established AI developers.
When the idea of “vibe coding” first emerged, its meaning revolved around informal, weekend projects that allowed beginners to explore coding with AI assistance. Public response was rapid, and the term became widespread, even earning recognition from dictionaries. However, previous news coverage focused more on the novelty and viral nature of the term, while recent developments showcase its impact on startup valuations and industry practices. Now, “agentic engineering” marks a more refined integration of AI into professional workflows, highlighting how much AI tools have matured in a short time.
How Does Agentic Engineering Differ from Vibe Coding?
Agentic engineering distinguishes itself from vibe coding by emphasizing intentional oversight and design in using AI agents. While vibe coding was often casual and experimental, the new framework describes a structured approach where engineers mainly delegate tasks to AI agents and supervise their outcomes. Karpathy clarified that “engineering” in this context refers to both the scientific and artistic aspects required to deploy AI agents effectively, focusing on output quality and reliability within professional settings.
Which Companies Are Influencing This Trend?
Startups such as Cursor, Lovable, and Replit have raised significant capital, indicating strong investor confidence in AI-assisted software development. Cursor’s fundraising efforts in late 2025 raised its valuation substantially, while Lovable and Replit have also secured investments that reflect a competitive ecosystem. Traditional AI leaders like OpenAI and Anthropic are enhancing their coding capabilities in response to the traction gained by these newer companies, creating a more dynamic marketplace for AI development tools.
What Are Industry Leaders Saying?
Karpathy, who has served both at OpenAI and Tesla, remains actively engaged in the AI field as a founder and investor. He commented on the shift from vibe coding to agentic engineering, stating,
“Directing and overseeing agents is now common practice, rather than writing code line by line.”
Addressing the skills involved, he said,
“There is an art and science to guiding AI agents so that software quality is not compromised.”
His ongoing support of startups engaged in agentic technologies underscores his belief in the model’s long-term significance to the software industry.
The conversation about agentic engineering versus vibe coding highlights critical changes in how software gets built. As more companies adopt agent-based workflows, engineers need to refine both management and technical skills to ensure productivity and code quality. Close attention to the growth of companies like Cursor, Lovable, and Replit can reveal broader trends in the application of large language models to coding, and investors’ keen interest shows that the commercial potential is still on the rise. Those working in software or aspiring to enter the field should watch developments in agentic engineering, as familiarity with AI-assisted frameworks could soon be a fundamental requirement for professional success.
