Meta Says Muse AI Agent Was Heavily Inspired by OpenClaw

This discussion did not start with Meta officially, but with observations from users. Ansh Nanda, co-founder of an AI app, posted a viral message saying Muse is essentially OpenClaw for mainstream users, and shared a conversation in which the agent described its own system files as 'matching' OpenClaw's. Other users then noticed that Muse also includes a SOUL.md file—a Markdown configuration used to define an agent's personality, values, and boundaries—with content nearly identical to OpenClaw's.

It is important to distinguish: Nanda's 'mainstream OpenClaw' is his personal judgment, while the near-identical SOUL.md content is an observation by other users. Both come from third parties, not from Meta's official confirmation. They created public pressure, but are not themselves the final ruling on product facts.

Facing the skepticism, Nat Friedman, product lead at Meta Superintelligence Labs and former GitHub CEO, gave the official line on X: Muse was built from scratch, but was deeply inspired by OpenClaw at the product level. This statement accommodates two seemingly contradictory claims at once—autonomous code, external inspiration.

As for why the same workspace file names were kept, Friedman's explanation was that the team believed OpenClaw creator Peter Steinberger had done these elements just right. He mentioned that Steinberger joined OpenAI earlier this year and called him a genius.

Friedman also revealed that after trying OpenClaw in January this year, he bought hundreds of Mac minis for the lab. He described Meta's goal as building a version that is safe, reliable, easy to use, and able to scale to billions of people. A Meta spokesperson pointed to Friedman's statement and did not add anything further.

The boundary of currently confirmable facts is quite clear: Meta acknowledges that Muse was deeply inspired by OpenClaw, and that Muse uses the same workspace file names as OpenClaw, including SOUL.md. But the claim of 'built from scratch' is currently supported only by Friedman's unilateral statement; the source contains no independent code-level comparison or third-party audit.

In other words, the similarities the outside world sees are concentrated at the level of product design and file naming, while the autonomy Meta claims is concentrated at the level of code implementation. The two do not necessarily conflict, but the existing material is insufficient to determine how deep the similarities actually extend.

It is worth noting that Friedman's evaluation of Steinberger, the purchase of hundreds of Mac minis, and the goal of 'scaling to billions of people' are all his own statements, belonging to source_claim rather than independently verified conclusions.

Muse recently reached No. 1 on the U.S. App Store, which means it has entered mainstream user view and is no longer just a lab project. At this point, any question about whether it is 'OpenClaw with a new skin' directly touches the product's originality narrative.

Friedman repeatedly emphasized 'safe, reliable, easy to use, scalable to billions of people,' which in effect draws a distinction between Muse and OpenClaw: OpenClaw is an open-source personal assistant project, while what Meta wants to build is a version scaled for the masses. This goal itself also constitutes a constraint—scale and safety requirements may be exactly why Meta is unwilling to simply copy an open-source implementation.

But conversely, if core design elements were indeed heavily borrowed, then how much pressure the phrase 'built from scratch' can withstand in public communication remains an unresolved question.

First, what is the actual degree of overlap between Muse and OpenClaw at the code level? The source provides no independent verification. Second, the 'nearly identical' SOUL.md content is a user observation; Meta did not directly respond to this specific allegation, only explaining the reason for keeping the file names. Third, whether the OpenClaw project side or Steinberger himself responded to Meta's statement is not recorded in the source.

In addition, the timing of Muse reaching No. 1 on the U.S. App Store is described as 'recently,' with no precise date given in the source; Friedman's trial of OpenClaw was in January, and the number of Mac minis purchased was 'hundreds,' both vague magnitudes rather than precise figures.

Before these gaps are filled, the more prudent reading is: this is a rare public statement in which an executive personally acknowledged inspiration from an open-source project, but the line between 'inspiration' and 'copying' is still drawn unilaterally by Meta.