The creator of the viral AI agent Clawdbot says he had to step back after becoming too obsessed with vibe coding.
Peter Steinberger, the developer behind Clawdbot — which later changed its name to Moltbot and is now known as OpenClaw — said in an episode of “Behind the Craft” podcast published Sunday that vibe coding pulled him into a “rabbit hole.”
“I was out with my friends and instead of, like, joining the conversation in the restaurant, I was just like, vibe coding on my phone,” he said.
“I decided, OK, I have to stop this more for my mental health than for anything else,” he added.
Clawdbot went viral last month in the tech community, attracting a wave of high-profile fans — from Y Combinator CEO Garry Tan to multiple partners at Andreessen Horowitz.
It is a personal AI agent designed to run continuously and plug into a wide range of consumer apps, including WhatsApp and Telegram. Users can ask the AI to manage their schedules, oversee vibe-coding sessions, and even create AI employees.
The AI agent has been widely praised and meme’d online, with some tech fans even buying Mac Minis specifically to run the AI, Business Insider’s Henry Chandonnet reported last week.
Steinberger said developers can fall into this trap of being hooked onto vibe coding, where building increasingly powerful AI tools creates the “illusion of making you more productive” without real progress.
Building new tools can feel rewarding and fun, but that can quietly blur into compulsion, he added.
With AI, developers can now “build everything,” but ideas and taste matter. Without them, developers risk building tools and workflows that don’t actually move a project forward, Steinberger said.
“If you don’t have a vision of what you’re going to build, it’s still going to be slop,” he added.
The hype around vibe coding
Vibe coding has continued to surge in popularity, with companies and developers promoting how AI can speed up software development.
Earlier this month, Anthropic said it built its new agentic work tool, Cowork, entirely using Claude.
“@claudeai wrote Cowork,” Anthropic’s product manager, Felix Rieseberg, wrote on X. “Us humans meet in-person to discuss foundational architectural and product decisions, but all of us devs manage anywhere between 3 to 8 Claude instances implementing features, fixing bugs, or researching potential solutions.”
Thanks to Claude, the agent came together quickly. “We sprinted at this for the last week and a half,” Rieseberg said during a livestream.
Still, despite the excitement around how fast vibe coding can produce new tools, tech leaders are warning that it has limits.
Google CEO Sundar Pichai said in November in a “Google for Developers” podcast interview that he won’t vibe code on “large codebases where you really have to get it right.”
“The security has to be there,” he added.
Boris Cherny, the engineer behind Anthropic’s Claude Code, said last month that vibe coding is great for prototypes or throwaway code, not software that sits at the core of a business.
”You want maintainable code sometimes. You want to be very thoughtful about every line sometimes,” he said in an episode of “The Peterman Podcast” published in December.
