Peter Steinberger built OpenClaw as a side project. A personal AI agent that connects to WhatsApp, Telegram, Slack, Discord, runs on your hardware, talks to your tools. Nothing particularly novel on paper.
9,000 GitHub stars. Nice for a side project.
Then something happened. Four months later: 250,000+ stars. More than React. NVIDIA publicly called it “to agentic AI what GPT was to chatbots.” In China, it went viral under the “raise a lobster” trend, with millions of users customizing their own agents.
Then Steinberger announced he is joining OpenAI. The project moves to a foundation.
What made it explode
The timing was perfect. People wanted AI agents. Not chatbots. Not assistants. Agents that DO things on their behalf. Send messages. Book appointments. Monitor systems. Act autonomously.
OpenClaw hit that demand at exactly the right moment with exactly the right distribution strategy: open source, runs locally, connects to the messaging apps people already use. No new app to download. No subscription. No cloud dependency.
The lobster meme in China (users treating their OpenClaw agents like virtual pets they “raise”) added the viral layer. Suddenly it was not just a developer tool. It was a cultural phenomenon.
The open source sustainability question
Steinberger leaving for OpenAI is the oldest story in open source. Creator builds something loved by millions. Big company hires the creator. Project moves to a “foundation.” Foundation either thrives (Linux) or slowly dies (too many to count).
The question is not whether OpenClaw is good technology. It is. The question is whether a 250,000-star project can survive without the person who made the architectural decisions that made it worth 250,000 stars.
History says: maybe. The odds are not great.
Why this matters for builders
If you are building anything in the AI agent space, OpenClaw is the benchmark now. Not for technology. For distribution.
Steinberger did not build the most sophisticated agent. He built the most accessible one. Runs locally. Connects to existing messaging. No subscription. That combination turns out to be worth more than technical sophistication.
The lesson for every builder: the best technology does not win. The most accessible technology wins. The most shareable technology wins. The technology that meets people where they already are wins.
250,000 engineers understood that in four months. How long will it take the rest of us?
Sources: DigitalOcean, Fortune