Andrej Karpathy sat in his new office at Anthropic and prepared to change the way the world builds software. The OpenAI founder and former Tesla AI director has defected to his former rival, bringing with him the concept of vibe coding. The method is simple: describe what you want in plain English and let the AI model do the work. Karpathy told his 2 million followers on X that the next few years at the frontier of AI would be especially formative.

But the engineers who actually built the machines are sounding the alarm. The creators of the OpenClaw AI agent warned this week that vibe coding is flooding the world with vibe slop. According to a report in the Wall Street Journal, the AI is generating massive amounts of low-value, dangerous code that contains hidden vulnerabilities. They warned that while the software might look like it works, the lack of human engineering oversight means eventually it will catch up to us.

The Cognitive Enclosure is tightening. As Google prepares to launch an Android version of its AI Studio, the ability to write code is being stripped from human experts and handed to black-box models. We are entering an era where the digital infrastructure of our lives is being built by amateurs prompting machines that they do not understand. The result is a world of slop where convenience is traded for systemic fragility.