OpenClaw Creator Urges AI Developers to Embrace Playfulness and Iterative Growth

Peter Steinberger, the creator behind the viral AI agent OpenClaw who now works at OpenAI, offers guidance for those diving into AI technology and agent development. His key takeaway? The most effective approach right now is to explore, experiment playfully, and avoid putting pressure on yourself to become an expert immediately.

“I wish I could claim I had a master plan from the start, but much of it was simply exploration,” Steinberger shared. “I envisioned tools that didn’t exist yet, and… well, I prompted them into reality.”

The developer discussed this during a conversation with Romain Huet, OpenAI’s Head of Developer Experience, on the debut episode of the company’s "Builders Unscripted" podcast. He reflected on OpenClaw's humble beginnings and his initial lack of a concrete roadmap.

Steinberger explained that he first built a tool for WhatsApp integration but temporarily shelved it to focus on other projects, assuming major AI labs would soon develop similar solutions.

“I was just experimenting a lot. My goal was essentially to have fun and inspire others,” Steinberger noted. However, by last November, he was surprised that no AI lab had built what he envisioned. This realization spurred him to create the first prototype of what became OpenClaw.

“The moment it truly clicked was during a weekend trip to Marrakesh. I found myself relying on it constantly because it was so convenient… The internet there wasn't great, but WhatsApp just works everywhere,” he said. The tool effortlessly helped him find restaurants, look up information, message friends, and more.

The more he engaged with the technology, the more Steinberger recognized the advanced problem-solving capabilities of modern AI models, drawing parallels to skilled coders.

“Now, they can essentially devise solutions independently, even without explicit programming,” he observed.

Throughout his development journey, Steinberger's workflow continuously improved. He emphasizes to fellow developers that this progression takes time and encourages persistence.

“There are developers who… write software the traditional way, but that approach is becoming obsolete,” he pointed out. He noted that some try AI-assisted ‘vibe coding’ but become frustrated with the initial results.

“I think ‘vibe coding’ is a bit of a misnomer,” Steinberger suggested, indicating the process isn't as effortless as the term implies. “People try AI but don't realize it's a skill to be developed,” he said, comparing learning to code with AI to learning an instrument like guitar.

“You won't be a guitar expert on day one,” he remarked. Instead, he advocates for a playful, exploratory mindset. Now, when crafting a prompt, he has an intuitive sense of how long it should take. If it takes longer, he analyzes what might have gone wrong and adjusts accordingly.

“My… advice is always to approach it playfully. Build something you've always wanted to build. If you have even a slight builder's mentality, there's likely an idea in the back of your mind. Just play with it.”

This spirit of experimentation and enjoyment is crucial, especially when many fear AI might replace their roles.

“If your core identity is ‘I want to create things, I want to solve problems,’ and if you're proactive and resourceful, you'll be in higher demand than ever,” Steinberger concluded.