Meta's Muse agent passes 2.5 million downloads as exec admits it was heavily inspired by OpenClaw

Start with the numbers, because the word hit has been worn thin and this time the figures are sturdier than the chatter. Per Quartz, Muse debuted on September 8 and quickly climbed to the number-one spot on the US App Store, and Meta stock rose 11% on Monday on the app's reception. Per Sensor Tower data cited by CBS News, Muse has topped 2.5 million downloads in two weeks, against 3.1 million for ChatGPT and 200,000 for Claude over the same period. CBS News also quotes John Mercer, head of global research at Coresight, who says that if Muse maintains momentum, it could maybe leapfrog Claude among consumers and get into the top tier of AI tools.

The second thread is the original source and wording of the admission, because this was not a press release but a public post. Per Quartz, Friedman wrote on Monday that Muse was definitely heavily inspired as a product by OpenClaw, saying that after using OpenClaw in January he purchased hundreds of Mac minis for the Meta Superintelligence Labs team so they could use it, and that lots of the team fell in love with using OpenClaw. His stated goal was to build something like OpenClaw that Meta could make safe and secure and easy to use and scale to billions of people. The report adds that the admission came after weeks of speculation among early Muse users: according to TechCrunch, the thread that set it off was posted by Ansh Nanda, an AI app co-founder, who claimed that Muse is literally OpenClaw for normies and shared a conversation in which Muse said the similarity between its system files and OpenClaw's was not a coincidence but a match.

The third angle is the technical detail, because the dispute here is specific rather than a vague similarity of ideas. Per Quartz, other users then discovered that Muse carries a SOUL.md file, the same type of plain-text document used in OpenClaw to govern an agent's character, guiding principles and operating limits, and according to TechCrunch the two files' contents were nearly indistinguishable. Asked why Muse used the same file names and a SOUL.md that so closely mirrored OpenClaw's, Friedman offered no pushback: he replied that the team thought Peter got those things exactly right. The report identifies Peter as Peter Steinberger, the Austrian developer who created OpenClaw and released it under an open-source license; per The Next Web, Friedman called Steinberger a genius whose harness was pioneering from the jump, and noted OpenAI hired Steinberger earlier this year.

The fourth thread is why this stayed in product territory rather than becoming a legal question, and how Meta responded. Per Quartz, because OpenClaw was released under an open-source license, the overlap in files has no clear legal dimension, leaving the debate squarely in product territory. The report adds that Meta issued no statement of its own, and when TechCrunch sought comment a spokesperson directed reporters to Friedman's post and declined to elaborate further. That handling is itself notable: under an open-source license, reuse at the code level is permitted, and what is genuinely scarce is the step of inventing the interaction paradigm in the first place. When a trillion-dollar company openly credits an open-source personal project as its inspiration, the industry conversation about who defines the agent form changes character.

The fifth thread is commercial friction, which says more about where agentic shopping is stuck than download counts do. Per CBS News, Amazon confirmed to the network that it has blocked the AI app from shopping on its e-commerce platform, citing security and user experience concerns; Quartz puts it more precisely, saying Amazon argued the agent conceals its identity and violates the platform's conditions of use. The reports also supply the demand-side backdrop: according to IBM, 41% of consumers use AI assistants to research products and 31% use them to search for deals. CBS News cites a Coresight survey of more than 1,000 consumers in which 30% said they would use an AI tool to compare prices for holiday shopping, with the same share saying they would use such tools to find deals or promo codes on holiday gifts. But Neil Saunders, a retail analyst at GlobalData Retail, cautions that people are still a little bit nervous about allowing AI agents to do the purchasing on their behalf.

Two things remain to watch. First, Muse is described as able to act on users' behalf, handling tasks such as booking a flight, decluttering inboxes and tracking price drops, which CBS News says is exactly what separates it from a traditional chatbot. That means its value depends less on how clever the model is and more on how many sites and accounts it can actually reach. Amazon's block points straight at the bottleneck: the ceiling is set by whoever cooperates least. Second, being heavily inspired is normal in the open-source world, but it shifts the scoreboard from who built it first to who can make it safer, more durable and available to more people, which per Quartz is the framing Friedman himself used. Every fact and figure here comes from Quartz, CBS News and the TechCrunch and The Next Web reporting they cite, with no speculation added.