Hyundai Autoever CEO turns coding instructor, warns against blind faith in AI

Hyundai Autoever CEO Ryu Seok-moon has stepped into the role of coding instructor. Personally demonstrating "vibe coding" — a development method in which users describe what they want in plain language and collaborate with AI to produce software — Ryu said the ability to use new technology efficiently matters far more than blind trust in it. He added that as Hyundai Motor Group accelerates its AI transformation, the group must weigh costs, productivity and data reliability to generate real business value.

Ryu led a vibe coding hands-on session for reporters Wednesday at Hyundai Motor Group's Yangjae headquarters in Seoul, held alongside the group's AX Achievement Showcase. The session followed the same vibe coding curriculum that Hyundai Motor Group Chairman Euisun Chung and other senior executives completed in the first half of this year.

Vibe coding lets developers describe a desired outcome in natural language and build the result in collaboration with AI, rather than writing code line by line. Because the same input can yield different outputs, prompt design and data reliability are critical. Through the session, Ryu said competitiveness in the AI era comes not from the technology itself but from using it efficiently.

"Saying you can do everything with vibe coding is too optimistic and premature, but dismissing it as useless is equally wrong," Ryu said. "You need to keep adjusting how you use it as the technology evolves, and you need a very solid understanding of software development."

The hands-on exercise used Claude to build a simple game through vibe coding. When participants typed in plain language — "make a game where you click on students dozing off in class to wake them up" — a working game was ready within minutes. Specific conditions were also set in natural language, such as failing if a student is not woken within two seconds of nodding off, with three failures ending the game.

Ryu explained that because large language models generate outputs based on probability, entering the same text does not always produce identical results. Even when a program appears to function normally on the surface, he said, such small variations can become serious defects in commercial or enterprise software.

Ryu said using AI does not automatically raise productivity. If a user loses focus while AI generates output, or stays stuck in a loop of repeatedly entering the same prompt, token costs can climb with no productivity gain to show for it.

"We use vibe coding and AI models to boost productivity, but a wrong approach to how you use them often ends up draining your attention entirely, with no productivity gain at all," Ryu said. "There are many cases where token costs skyrocket and the organization sees no return on its investment."

As Hyundai Motor Group's AI usage scales up rapidly, cost management has emerged as a key challenge. About 90 percent of the group's developer workforce now uses AI on the job, Ryu said, and he expects that share to rise further as security restrictions are eased.

Against that backdrop, the group is focused on maximizing results at minimum cost. Jin Eun-sook, president in charge of ICT at Hyundai Motor and Kia, said executives also need "a sense of what problems AI investment will solve and how to realize its benefits."

Chairman Chung and other senior executives have placed emphasis on building the capacity to understand and evaluate AI firsthand. Chung expressed a desire to try vibe coding himself and joined the training with executives in the first half of this year. He proposed the sessions on the grounds that executives must properly understand AI technology before they can factor it into decision-making.

"Chairman Chung stressed that it is important to clearly know what AI can and cannot do, and where it can and cannot be applied — understanding both the technology's limits and its potential," Jin said. "He urged all executives to make good use of AI and apply it appropriately where it is needed."