On February 2, 2025, Andrej Karpathy, a co-founder of OpenAI and former director of AI at Tesla, posted on X defining what he called "a new kind of coding"4. He described using Cursor Composer with voice commands, barely touching his keyboard, and clicking "Accept All" on every AI suggestion without looking at diffs. When errors popped up, he'd copy-paste them back into the AI for fixes. His codebase grew in ways he didn't fully understand, but he could still get a working application by "poking and prodding" and sometimes requesting random changes until bugs disappeared1.

Karpathy acknowledged the limits: "It's not too bad for throwaway weekend projects... I just see things, say things, run things, and copy-paste things, and it mostly works"1. The post was self-aware and partly tongue-in-cheek, an expert programmer intentionally not using his skills to test how far AI could carry a project. The post picked up over 27,000 likes within a month4.

Vibe coding moved from internet joke to mainstream tech discourse faster than almost any developer meme before it. Coverage in The New York Times, The Guardian, and Ars Technica brought the concept to non-technical audiences1. Enterprise consulting firms like Thoughtworks dedicated podcast episodes and internal strategy discussions to it2.

The Y Combinator connection gave the term institutional weight. Reports that a quarter of the Winter 2025 batch relied on heavily AI-generated codebases made vibe coding feel like a real industry shift rather than just a Twitter joke4. College professors and software managers began sharing stories about dealing with young developers who couldn't debug their own AI-generated code, adding an intergenerational dimension to the discourse4.

Rick Rubin's involvement pulled vibe coding into the art-and-creativity conversation. His book "The Way of Code" treated the concept as part of a longer philosophical tradition, comparing AI-assisted creation to musical improvisation and remix culture6. The a16z podcast framing it as "punk rock" gave Silicon Valley's venture class a narrative they could champion6.

The cybersecurity community raised alarms. Experts pointed out that vibe-coded applications could introduce subtle bugs, security vulnerabilities, and unmaintainable complexity at scale1. The gap between "it works on my machine" and "it's safe to deploy" became a recurring theme in professional discussions about AI-assisted development2.

The speed of vibe coding's semantic diffusion was unusual even by internet standards. Birgitta Böckeler, AI for Software Delivery Lead at Thoughtworks, noted that the term went from a single tweet to dozens of competing definitions in under seven weeks2. Some developers stripped it down to "you use an AI coding tool and don't review anything it does," while others stretched it to cover any use of chat-driven development, including the supervised agentic modes emerging in tools like Cursor, Cline, and Windsurf2.

The corporate world took notice fast. Thoughtworks reported that a senior leader at one of their large client accounts proposed using vibe coding to reduce team size and cut costs, treating the meme as a legitimate methodology2. This forced consultants to ask uncomfortable clarifying questions about what the client actually meant. In many cases, executives had simply discovered agentic coding assistants through Karpathy's viral post and wanted to apply them broadly, without understanding the distinction between throwaway weekend projects and production software2.

The cultural fault lines ran deep. One camp celebrated vibe coding as democratization, pointing to the Y Combinator Winter 2025 batch where 25 percent of startups relied on 95 percent AI-generated code4. The other camp saw it as a disaster waiting to happen. Security experts like Lilly Ryan at Thoughtworks flagged the risks of deploying code that no human had actually read, warning about subtle bugs, security flaws, and unmaintainable complexity2. The Substack article "Vibe Coding: Revolution or Reckless Abandon?" argued that developers needed to act as "expert curators and critical validators" rather than passive prompt-givers1.

The meme crossed into unexpected territory when legendary music producer Rick Rubin embraced it. In a conversation on the a16z podcast "The Ben & Marc Show," Rubin discussed writing "The Way of Code," a book that blended the 3,000-year-old Tao Te Ching with AI and vibe coding philosophy6. Marc Andreessen and Ben Horowitz explored how Rubin went from becoming a vibe coding meme to producing a "creative manifesto" for the AI age, with Rubin calling the movement "the punk rock of software"6.

Software blogger Steve Yegge took things further. He became a vocal advocate for vibe coding and coauthored a book titled "Vibe Coding" with Gene Kim7. In January 2026, Yegge launched Gas Town, a vibe coding orchestrator that he described as "100% vibe coded. I've never seen the code, and I never care to, which might give you pause"7. An unknown person created a memecoin called $GAS tied to the project, paying transaction fees to Yegge. He endorsed it publicly. The coin collapsed in what appeared to be a rugpull, though Yegge made over $290,000 from the venture before it imploded7.

Meanwhile, the broader cultural reading of vibe coding gained traction. Writers framed it as a rejection of hyper-structured workflows, SCRUM ceremonies, and ticket-driven labor3. The original aesthetic included RGB-lit setups, lo-fi beats, and captions like "No Jira, no deadlines, just vibes"3. In educational spaces, teachers began experimenting with vibe coding approaches to lower barriers to entry for students intimidated by traditional programming3. The concept touched something real about developer burnout and the desire to reconnect with the creative, improvisational side of building software3.