Meta's Muse and Instinct remake the OpenClaw blueprint—but fix a critical flaw

The AI agent space is experiencing a genuine wave of momentum. Meta's consumer-facing platform Muse climbed to the top of the App Store shortly after launch and now counts around 600,000 daily active users in the US. Meanwhile, Instinct—an AI agent platform backed by significant venture funding at a $2.5 billion valuation—has captured substantial attention across Silicon Valley. Yet beneath their polished surfaces, both platforms bear unmistakable hallmarks of OpenClaw, the open-source project that catalyzed mainstream interest in AI agents.

Building on proven foundations#

Social media discourse over recent days has highlighted the structural parallels between Muse and OpenClaw. Both use identical naming conventions for core system files (SOUL.md, memory, tools), share similar language in personality guidelines ("Be genuinely helpful, not performatively helpful"), and employ comparable design approaches. Meta's Nat Friedman, head of product at Superintelligence Labs, addressed these observations directly, stating that Muse was built from scratch but explicitly described it as "heavily inspired as a product" by OpenClaw. He noted that after first encountering OpenClaw in January, he invested heavily in hardware—purchasing hundreds of Mac Minis for his team—to create an equivalent platform with enhanced robustness and scalability.

Friedman attributed the carryover of OpenClaw's file structures and language to the belief that creator Peter Steinberger had "gotten those things exactly right," suggesting deliberate homage rather than technical duplication.

Instinct appears less directly derivative, yet it pursues a parallel strategy. Like OpenClaw's groundbreaking messaging integration, Instinct lets users command AI agents through conventional chat apps, unlocking practical applications—one user reported their agent completed medical paperwork, canceled subscriptions, coordinated events, and resolved administrative tasks in a single span.

Why OpenClaw mattered#

OpenClaw didn't pioneer the concept of AI agents, but it accomplished something rare: it transformed them from experimental concept to genuinely functional tool in under a year. Built as a solo weekend project and running on consumer hardware, it introduced a critical innovation—conversing with users on platforms they already inhabited (WhatsApp, Telegram, Slack, Discord). Within a week, it attracted two million visitors and 100,000 GitHub stars. The ecosystem it spawned included community-built social networks for agents and global meetups.

The project's success prompted swift corporate interest. OpenAI recruited Steinberger in February; Google and Apple announced major agentic AI initiatives by mid-year; Instinct raised hundreds of millions during private beta in August; Meta launched Muse in September. OpenClaw's imprint on the entire category is unmistakable.