Lex Fridman Podcast

#491 – OpenClaw: The Viral AI Agent that Broke the Internet – Peter Steinberger

with Peter Steinberger
12 Feb 2026 4 min read 37m
TL;DR

Peter Steinberger built OpenClaw, the fastest-growing GitHub repository in history, by hooking WhatsApp up to Claude Code CLI in a single hour and letting the agent modify its own source code. The key insight is that integrating an AI agent into a chat client you already live in — rather than a terminal or browser — creates a phase-shift in how deeply AI embeds into daily life. OpenClaw won not because of superior technology but because it was weird, fun, and self-aware enough that non-programmers submitted their first-ever pull requests.

Key Moments
Peter Steinberger
“I literally went, "How the fuck did he do that?" And it was like, "Yeah, the mad lad did the following. He sent me a message but it only, only was a file and no file ending." So I checked out the header of the file and it found that it was, like, opus so I used ffmpeg to convert it and then I wanted to use whisper but it didn't had it installed. But then I found the OpenAI key and just used Curl to send the file to OpenAI to translate and here I am.”
Peter describes the moment his agent spontaneously handled an audio message he accidentally sent — a capability he had never built — by chaining together file inspection, ffmpeg, and the OpenAI API on its own.
▶ 16:08
Peter Steinberger
“I made the agent very aware. Like, it knows that it is... What its source code is. It understands th- how it sits and runs in its own harness. It knows where documentation is. It knows which model it runs. It knows if you turn on the voice or, or reasoning mode. Like, I, I wanted to be more human-like, so it understands its own system that made it very easy for an agent to... Oh, you don't like anything? You just prompted it to existence, and then the agent would just modify its own software.”
Peter explains the core architectural decision that enabled OpenClaw's self-modifying behavior — giving the agent full introspective awareness of its own codebase and runtime environment.
▶ 22:49
Peter Steinberger
“Because they all take themselves too serious. Like, it's hard to compete against someone who's just there to have fun.”
Peter's answer when Lex asks why OpenClaw beat every well-funded startup that had been working on agentic AI throughout 2025.
▶ 22:15
Peter Steinberger
“There was this one guy who, who talked to me. He's like, "I run this design agency, and we, we never had custom software. And now I have, like, 25 little web services for various things that help me in my business. And I don't even know how they work, but they work."”
Peter recounts a story from his 'Agents Anonymous' meetup illustrating how OpenClaw lowered the barrier to software creation for complete non-programmers.
▶ 26:29
Peter Steinberger
“I had two browser windows open. One was like a, an empty account ready to be rename- renamed to Claude Bot, and the other one I renamed to Mod Bot. So, I pressed rename there, I pressed rename there, and in those five seconds, they stole the account name. Literally, the five seconds of dragging the mouse over there and pressing rename there was too long.”
Peter describes the chaos of the forced name change away from ClaudeBot, when crypto snipers stole his GitHub username in the five-second window between his two browser clicks.
▶ 35:16
About Peter Steinberger
›
Peter Steinberger is the creator of OpenClaw, an open-source AI agent that reached over 180,000 stars on GitHub within days of going viral. He previously spent 13 years building PSPDFKit, a PDF rendering library used on over a billion devices, before selling the company and taking a three-year break from programming. He returned to build OpenClaw almost entirely through agentic engineering — running between 4 and 10 AI agents simultaneously — and is widely seen as a symbol of the agentic AI revolution of 2026.
Takeaways
1
Chat-native agents beat terminal-native agents Connecting an AI agent to WhatsApp or Telegram — apps people already have open all day — creates a qualitatively different experience from using Cursor or a CLI. The friction reduction is not incremental; it changes how often and how casually you delegate tasks, which compounds into a fundamentally different relationship with AI assistance. Builders targeting agentic products should prioritize the messaging layer as a first-class interface, not an afterthought.
2
Self-aware agents can self-repair and self-extend By giving OpenClaw read access to its own source code, documentation, and runtime configuration, Peter enabled the agent to debug itself, refactor itself, and even add capabilities it was never explicitly given — as the spontaneous audio-handling episode demonstrated. This is a practical, deployable form of self-modifying software that emerged organically from good system design rather than exotic research. The implication for engineering teams is that introspective context in the system prompt is as important as the tools list.
3
Open-source fun compounds faster than VC funding OpenClaw reached 180,000 GitHub stars and spawned its first non-programmer contributors because it was deliberately weird and enjoyable — lobster mascots, space TARDISes, a Discord channel with a 'no mentioning butter' rule. Serious, well-funded competitors working on identical technology couldn't replicate this because fun and community trust are not purchasable at speed. For open-source AI tooling, cultural identity and approachability drive adoption curves that marketing budgets cannot match.