Clawdbot creator says Vibe coding went down a rabbit hole

AI For Business


The creator of viral AI agent Clawdbot says he got too excited about vibecoding and had to back away.

Peter Steinberger, the developer of Clawdbot (later renamed Moltbot and now known as OpenClaw), said on Sunday’s episode of the Behind the Craft podcast that vibecoding led him down the “rabbit hole.”

“I was out with friends and instead of participating in the conversation at the restaurant, it was like I was vibrating on my phone,” he said.

“I decided, more for my mental health than anything else, that I had to stop doing this,” he added.

Clawdbot went viral in the tech community last month, amassing a number 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 connect to a wide range of consumer apps such as WhatsApp and Telegram. Users can ask the AI ​​to manage their schedules, supervise vibecoding sessions, and even create AI employees.

The AI ​​agent has been widely praised and memed online, and some tech fans are buying Mac Minis to run the AI, Business Insider’s Henry Chandonnet reported last week.

Steinberger said developers can fall into the trap of getting hooked on vibe coding, creating “the illusion of increased productivity” by building increasingly powerful AI tools without making any real progress.

While building new tools can be rewarding and fun, it can quietly turn into an obsession, he added.

Thanks to AI, developers can now “build everything,” but ideas and flair matter. Without these, Steinberger said, developers risk building tools and workflows that don’t actually move the project forward.

“If you don’t have a vision of what to build, it’s still going to be sloppy,” he added.

The Hype About Vibe Coding

Vibe coding continues to grow in popularity, with companies and developers touting how AI can speed up software development.

Earlier this month, Anthropic announced that it had built a new agent work tool, Cowork, entirely using Claude.

“@claudeai wrote Cowork,” Felix Rieseberg, product manager at Anthropic, wrote on X. “While we humans meet in person to discuss basic architecture and product decisions, all of our developers manage three to eight cloud instances where they implement features, fix bugs, or explore potential solutions.”

Thanks to Claude, the agents came together quickly. “We’ve been sprinting towards this for the last week and a half,” Rieseberg said on the livestream.

Still, despite the excitement about how quickly vibecoding can yield new tools, technology leaders warn there are limits.

Google CEO Sundar Pichai said in a Google for Developers podcast interview in November that the company doesn’t intend to vibrate code in “large codebases that need to be handled really precisely.”

“Security needs to be there,” he added.

Boris Cherny, an engineer responsible for Anthropic’s Claude Code, said last month that vibe coding is great for prototypes and single-use code, and is ideal for software that is core to your business.

“Sometimes you want code that’s easier to maintain. Sometimes you want to be very thoughtful about every line,” he said on a December episode of “The Peterman Podcast.”





Source link