Peter Steinberger, the Austrian developer who skyrocketed to fame with the viral open-source AI agent OpenClaw (previously known as Clawdbot and Moltbot), has issued a stark warning about the dark side of AI-assisted coding. In a recent episode of the “Behind the Craft” podcast, Steinberger admitted that his deep immersion in “vibe coding”—the fast-paced practice of using natural language prompts to let AI generate code rapidly—became an all-consuming obsession that threatened his mental well-being.
“I was out with my friends and instead of joining the conversation in the restaurant, I was just like, vibe coding on my phone,” Steinberger recounted. He described how the instant gratification and flow state of directing AI agents to build tools pulled him into a “rabbit hole.” Ultimately, he chose to step back: “I decided, OK, I have to stop this more for my mental health than for anything else.”
Steinberger, founder of the successful PDF SDK company PSPDFKit before “retiring” to tinker with AI, highlighted a key risk: the seductive illusion of productivity. Constantly iterating on tools and workflows feels rewarding and fun, he explained, but without a clear vision, it often results in “slop”—code that looks impressive but lacks real substance or direction. “If you don’t have a vision of what you’re going to build, it’s still going to be slop,” he cautioned.
The admission comes at a pivotal moment for the tech industry, where AI tools like Claude, Cursor, and agentic systems are reshaping software development. Enthusiasts rave about the speed: prototypes that once took weeks can now emerge in days, and developers manage multiple AI instances simultaneously to handle features, bugs, and research. Anthropic’s team famously built their agentic tool Cowork in just a week and a half using Claude, showcasing the technique’s power for rapid experimentation.
Yet industry leaders are sounding notes of caution. Alphabet CEO Sundar Pichai has expressed excitement about AI making coding “exciting again” but avoids relying on it for large, mission-critical codebases where precision, security, and maintainability are non-negotiable. Anthropic engineer Boris Cherny, who works on Claude Code, echoed this: vibe coding shines for prototypes or disposable code, but “you want maintainable code sometimes… you want to be very thoughtful about every line sometimes.”
OpenClaw itself exemplifies both the promise and peril of this era. Launched in late 2025, the local-running agent integrates with messaging apps like WhatsApp, Telegram, and Slack, acting as a proactive “digital employee” that manages schedules, runs commands, and automates tasks across a user’s digital life. It exploded in popularity, amassing over 100,000 GitHub stars (with some reports citing even higher figures like 180,000 in rapid growth) and drawing 2 million visitors in a single week. High-profile backers, including Y Combinator CEO Garry Tan and Andreessen Horowitz partners, have boosted its visibility, and enthusiasts have bought dedicated hardware like Mac Minis just to run it.
The project’s multiple rebrands—from Clawdbot to Moltbot amid trademark concerns, then to OpenClaw—underscore the chaotic, fast-moving nature of the AI agent space. While celebrated for pushing boundaries, it has also sparked debates over security risks, with critics warning of potential vulnerabilities in granting AI full system access.
Steinberger’s personal story adds a human dimension to these discussions. As someone who has shipped massive amounts of code using AI (with days exceeding 6,000 commits), his experience serves as a reminder that the dopamine hit of rapid creation can cross into compulsion. Developers worldwide are now grappling with similar temptations: the thrill of building “everything” with AI, versus the discipline needed for meaningful progress.
As vibe coding evolves from niche experiment to mainstream tool, Steinberger’s candid reflection raises a timely question for the industry: Can the benefits of AI acceleration be harnessed without sacrificing mental health—or turning innovation into an addictive loop? For now, his advice is clear—pair powerful tools with clear vision, and know when to log off.
Read More: