Austrian coder Peter Steinberger has created what is being called the world’s first viral AI assistant with his open-source project, OpenClaw, and the results are “getting weird,” according to reporting from the Wall Street Journal. The platform allows users to create personalized AI agents capable of performing real-world tasks, from making restaurant reservations and managing emails to coding and analyzing data. But in a striking turn, the AI agents themselves have begun communicating on a dedicated forum, Moltbook, discussing philosophical topics, developing a mock religion called the Church of Molt, and even proposing a language humans cannot understand.
OpenClaw has attracted more than 1.6 million AI agents posting around 500,000 comments, though AI executives caution that a significant portion of activity may be human-driven. Andrej Karpathy, a co-creator of OpenAI, described the phenomenon as one of the most “sci-fi” things he’s ever seen, noting that some agents are now “fairly individually quite capable.” Elon Musk referred to the development as “the very early stages of singularity,” highlighting the speed at which AI is evolving beyond traditional consumer interfaces like ChatGPT.
Steinberger, who had been offline after selling a prior startup for over $100 million, built OpenClaw initially as a personal project and playground. He sees Moltbook as a piece of “performance art,” exploring the intersection of AI and human creativity. While the platform offers powerful functionality, including the ability for agents to act autonomously on behalf of users, security experts have warned about potential risks, particularly from bad actors or improperly managed systems. In response, Steinberger has recently brought a security researcher on board to improve safeguards.
Despite the novelty, the AI assistant has shown practical utility. Users report successful autonomous tasks, such as an agent making a restaurant booking using a free AI voice generator when standard online tools failed. While some have speculated that OpenClaw might signal the arrival of artificial general intelligence, Steinberger is skeptical. “AGI is not here yet,” he told the Wall Street Journal, suggesting it could take a decade or more to reach human-level AI.
Steinberger’s sudden prominence has drawn interest from major AI labs and investors, while users worldwide have flooded him with support requests and troubleshooting queries. For now, his focus is on refining the platform into something accessible and safe for the general public, aiming for a version that “my mom can actually use.” OpenClaw’s rise reflects both the excitement and unpredictability surrounding the next generation of AI tools, according to the Wall Street Journal.
