Cardano Founder Reacts as OpenAI Co-Founder Promotes “Vibe Coding”

Cardano founder Charles Hoskinson reacts to the promotion of “vibe coding” by an OpenAI co-founder.

While the significant breakthroughs in generative artificial intelligence in recent years have sparked excitement, they have also triggered equal levels of concern over humanity’s growing reliance on the tool.

In the latest instance highlighting this, OpenAI co-founder Andrej Karpathy has talked up a concept called “vibe coding,” eliciting several concerned reactions, including from Cardano founder Charles Hoskinson.

“There’s a new kind of coding I call ‘vibe coding,’ where you fully give in to the vibes, embrace exponentials, and forget that the code even exists,” Karpathy wrote in an X post on Monday, February 3.

The computer scientist is referring to a type of coding where all the work, including debugging, is left to a large language learning model.

“I barely touch the keyboard,” Karpathy wrote, adding that he uses AI dictation apps like SuperWhisper to guide the LLM.

Karpathy also admitted to accepting all code changes from the LLM without a second thought. He asserted that the approach has largely worked for him and is helpful in “throwaway” projects.

Reacting to Karpathy’s post, Cardano’s Charles Hoskinson highlighted potential problems this approach could pose, sharing a GIF showcasing the melting face scene from “Raiders of the Lost Ark.”

My QA Team https://t.co/AXvWDd6h07 pic.twitter.com/GlreZWBeAU

— Charles Hoskinson (@IOHK_Charles) February 3, 2025

Expressing similar sentiments, prominent programmer Casey Muratori asserted that the approach was likely going to make code worse.

These views come from a lack of insight into the processes and reasoning behind the AI’s coding decisions. This can make the code difficult to maintain, potentially leading to financial costs.

But while over-reliance on AI in coding likely leads to too many unknowns, developers recently seem to agree that the technology holds potential as a tool to help with the often arduous and frustrating process of debugging by pointing out likely issues with code.

Still, even in this instance, some caution against relying on the AI to fix the code after a likely issue has been spotted.

Source