Meta Admits Muse’s Likeness to OpenClaw Isn’t a Coincidence: Inside the Agent Playbook

1. The Viral Tweet That Started It All

When Meta rolled out Muse, early testers immediately noticed something unusual. Most enterprise chatbots—from ChatGPT to Claude to Gemini—present a polished, opaque conversational interface. But Muse behaved differently: it had a distinctly hacker-centric personality, structured memory directories, and a transparent workspace model.

The dam broke when AI founder Ansh Nanda posted an explosive interaction with Muse on X:

“Muse is LITERALLY OpenClaw for normies.”

In Nanda’s posted screenshots, an engineer asked Muse why its internal workspace mirrored Peter Steinberger’s open-source agent framework, OpenClaw. In an astonishingly candid response, Muse admitted that the resemblance between its own configuration and OpenClaw was not just “a coincidence” but rather “a direct architectural match.”

2. What Is SOUL.md? The Secret Sauce of Personal Agents

Within hours, developers on X and GitHub began digging into Muse's file system prompts. The most striking discovery was the presence of a file named SOUL.md:

In traditional LLM system prompts, developers cram personality rules, negative constraints, and safety guidelines into one monolithic text block. Peter Steinberger (founder of PSPDFKit and creator of OpenClaw) pioneered a modular, file-system-first architecture where the agent reads plain markdown files:

A third developer noted that not only were the files named identically, but the core guidance directives inside Muse's SOUL.md—defining voice warmth, conciseness, proactive execution, and error handling—were copied nearly word-for-word.

3. Nat Friedman’s Response: "Peter Got Those Things Exactly Right"

Rather than issuing a standard PR denial, Nat Friedman—former CEO of GitHub who joined Meta to steer AI products alongside Chief AI Officer Alexandr Wang—responded with total candor:

“We built Muse from scratch, but it is definitely heavily inspired as a product by OpenClaw. After I used OpenClaw in January, I bought hundreds of Mac minis for the MSL team, and lots of us fell in love with using OpenClaw (and other personal agents).”

“@steipete [Peter Steinberger] is a genius, and his harness was pioneering from the jump. I think a lot of people were inspired by it. Our goal with Muse was to build something like OpenClaw that we could make safe and secure and easy to use and scale to billions of people.”

4. The Open Source Paradox: From Indie Hack to Corporate Powerhouse

The story of OpenClaw and Muse highlights the unique dynamics of AI development in 2026. Peter Steinberger released OpenClaw as an open-source passion project. Its breakthrough harness showed how giving an LLM direct access to local files and a well-crafted personality could create an agent that felt like a real autonomous partner rather than a bland text predictor.

The momentum was so intense that OpenAI acquired Steinberger earlier this year to help architect its own next-generation operator tools. Meanwhile, Meta did what Meta has historically done better than anyone else: recognize a grassroots product triumph, adopt its foundational design language, and mobilize world-class engineering infrastructure to bring it to hundreds of millions of users.

Muse recently reached No. 1 on the U.S. App Store, with analytics indicating its early mobile adoption velocity has outpaced even the explosive launch of ChatGPT.

5. Key Takeaways for AI Engineers and Founders

• Adopt Modular Personality Files: Ditch monolithic hardcoded system strings. Use structured markdown files like SOUL.md that can be versioned, edited, and dynamically adapted.
• Embrace Local Sandboxing: Developing agents with rapid local feedback loops (like MSL's Mac mini clusters) drastically accelerates engineering velocity compared to cloud-only debugging.
• Open Source Leads the Frontier: The best consumer product ideas are not born in corporate committee rooms—they emerge from passionate indie builders testing edge ideas in public.