Presenters
Source
🚀 The State of the Claw: Navigating the Future of Agentic AI 🤖
In just five months, OpenClaw has achieved what many thought impossible: it has become the fastest-growing open-source project in GitHub’s history. With nearly 30,000 commits, 2,000 contributors, and a trajectory that defies traditional growth charts, OpenClaw is redefining how we interact with personal AI agents.
Peter Steinberger, the project’s creator and a recent addition to the OpenAI team, took the stage to discuss the “State of the Claw,” the challenges of hyper-growth, and why the future of AI belongs to the users.
🛠️ The “Stripper Pole” Growth & Its Challenges
Growth this rapid comes with a unique set of growing pains. Steinberger describes managing the OpenClaw Foundation as “running a company on hard mode.” With volunteers driving the bus, the primary challenge is the bus factor—ensuring the project doesn’t grind to a halt if key contributors step away.
To scale, Steinberger has brought in heavyweight support:
- Industry Collaboration: Engineers from Nvidia, Microsoft, Red Hat, Tencent, and ByteDance are now actively contributing, focusing on everything from security and dockerization to MS Teams integration.
- The Foundation: To ensure the project remains neutral and sustainable, Steinberger is establishing the OpenClaw Foundation, a move inspired by the success of other open-source legends.
🛡️ Security: The “Insecure by Default” Narrative
OpenClaw has faced intense scrutiny, with 1,142 security advisories reported to date—a staggering 16.6 per day. Steinberger addresses these critiques head-on:
- The “Slop” Factor: Many reports are generated by automated agents looking for “credits” in the security community. Steinberger notes that the higher the criticality score, the more likely the report is to be “slop”—theoretical exploits that don’t apply to real-world usage.
- The Reality of Risk: Most “critical” vulnerabilities (like the CVSS 10.0 issue that panicked some) require users to actively fight the recommended security setup to even be exposed.
- The Lesson: “If you don’t play by these rules, you can get some really fun interactions,” Steinberger quips. He emphasizes that users must follow security documentation—such as sandboxing agents in group chats and avoiding sudo mode—to maintain a secure environment.
🦾 The Future: Ubiquitous Agents & “Dreaming”
Steinberger envisions a future where AI isn’t just on your phone, but ubiquitous.
- Ubiquitous Computing: He wants to talk to his agent from any room in his house, with the agent intelligently using nearby displays (like iPads) to visualize information.
- Dreaming: Inspired by human biology, the project is exploring “dreaming”—a process where agents perform “garbage collection” on memory logs during idle time to consolidate long-term knowledge.
- The Human Edge: Despite the power of AI, Steinberger argues that taste and system design remain the ultimate moats. “The way to the mountain is never a straight line,” he says. An agent can execute tasks, but it lacks the human intuition to know which tasks are worth pursuing.
🎙️ Q&A Highlights: Insights from the Stage
In a moderated session with Swix, Steinberger tackled the tough questions:
- On OpenAI’s Role: Steinberger clarifies that OpenAI hasn’t “bought” OpenClaw. Instead, the company understands that for AI to flourish, users need to experiment with agents in an open, customizable environment. “Everybody in the industry wins if more people spend time with AI,” he says.
- On Coding Workflow: Steinberger admits to running multiple parallel agent sessions to maintain velocity. While some advocate for “dark factory” (fully automated coding), he insists that iterative, human-guided development is still essential for building software with “soul.”
- On Essential Skills: For future engineers, Steinberger recommends focusing on system design and the power of saying no. Because AI makes it trivial to generate code, the challenge has shifted from how to build to what to build and how to keep the system maintainable.
🎯 The Bottom Line
OpenClaw is more than just code; it is a movement toward data sovereignty. By keeping the agentic stack open and local-first, Steinberger is building a future where you—not a massive corporation—own the keys to your digital life.
As Steinberger puts it: “The more powerful you make it, the more it can do for you, but the more you also have to understand what it does.” 🌐✨