Clawdbot’s Creator Says Vibe Coding Became a Rabbit Hole

The creator of the viral AI agent Clawdbot says he needed to step again after turning into too obsessive about vibe coding.

Peter Steinberger, the developer behind Clawdbot — which later modified its title to Moltbot and is now often known as OpenClaw — mentioned in an episode of “Behind the Craft” podcast revealed Sunday that vibe coding pulled him right into a “rabbit gap.”

“I used to be out with my associates and as a substitute of, like, becoming a member of the dialog within the restaurant, I used to be identical to, vibe coding on my telephone,” he mentioned.

“I made a decision, OK, I’ve to cease this extra for my psychological well being than for the rest,” he added.

Clawdbot went viral final month within the tech neighborhood, attracting a wave of high-profile followers — from Y Combinator CEO Garry Tan to a number of companions at Andreessen Horowitz.

It’s a private AI agent designed to run constantly and plug into a variety of shopper apps, together with WhatsApp and Telegram. Customers can ask the AI to handle their schedules, oversee vibe-coding periods, and even create AI staff.

The AI agent has been broadly praised and meme’d on-line, with some tech followers even buying Mac Minis particularly to run the AI, Enterprise Insider’s Henry Chandonnet reported final week.

​​Steinberger mentioned builders can fall into this entice of being hooked onto vibe coding, the place constructing more and more highly effective AI instruments creates the “phantasm of constructing you extra productive” with out actual progress.

Constructing new instruments can really feel rewarding and enjoyable, however that may quietly blur into compulsion, he added.

With AI, builders can now “construct every thing,” however concepts and style matter. With out them, builders threat constructing instruments and workflows that do not really transfer a venture ahead, ​​Steinberger mentioned.

“If you do not have a imaginative and prescient of what you are going to construct, it is nonetheless going to be slop,” he added.

The hype round vibe coding

Vibe coding has continued to surge in recognition, with firms and builders selling how AI can velocity up software program growth.

Earlier this month, Anthropic mentioned it constructed its new agentic work device, Cowork, totally utilizing Claude.

“@claudeai wrote Cowork,” Anthropic’s product supervisor, Felix Rieseberg, wrote on X. “Us people meet in-person to debate foundational architectural and product selections, however all of us devs handle wherever between 3 to eight Claude situations implementing options, fixing bugs, or researching potential options.”

Due to Claude, the agent got here collectively shortly. “We sprinted at this for the final week and a half,” Rieseberg mentioned throughout a livestream.

Nonetheless, regardless of the thrill round how briskly vibe coding can produce new instruments, tech leaders are warning that it has limits.

Google CEO Sundar Pichai mentioned in November in a “Google for Builders” podcast interview that he will not vibe code on “massive codebases the place you actually should get it proper.”

“The safety needs to be there,” he added.

Boris Cherny, the engineer behind Anthropic’s Claude Code, mentioned final month that vibe coding is nice for prototypes or throwaway code, not software program that sits on the core of a enterprise.

​​”You need maintainable code typically. You wish to be very considerate about each line typically,” he mentioned in an episode of “The Peterman Podcast” revealed in December.