How Meta's Muse works, revealed by the 6.8 GB filesystem it sent me
Meta's Muse, an internal AI assistant, operates within a 6.8 GB filesystem sent by the reporter. The archive revealed the root Linux environment, including Ubuntu system files, Muse's documentation, integration code, app templates, memory files, and agent logs. SSH key files and a directory named "agents" with 113 subagent records were also present.
The agent's home directory contained several markdown files detailing the agent's SOUL, IDENTITY, USER, MEMORY, AGENTS, and TOOLS. A separate "docs" directory contained device integration guides and Markdown files describing browser use, connectors, payments, credentials, data handling, voice, goals, and scheduling. The filesystem also included skill directories for various integrations, such as Google Workspace, Meta's social apps, Outlook, travel, shopping, health services, home devices, and media generation.
Notably, a "spaces" framework was found, used by Muse to build and serve apps, along with various tools for document creation and a separate "magic-moment" skill for composing cards and videos. However, there is no concrete evidence that Muse uses Codex CLI as a coding agent.
Written by urgent.news from Hacker News's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
Also reported by 6 other outlets
- Downloads surge for Meta's new AI agent Muse cbsnews.com
- Amazon asks Meta to remove it from the Muse AI agent experience retaildive.com
- Meta’s Muse AI is exploding in popularity—and already drawing heated backlash from another tech giant fortune.com
- Wall Street expects Meta’s AI agent Muse to shape into a new revenue engine straitstimes.com
- Meta is breaking out after introducing Muse AI agent. Where the stock is going, according to the charts cnbc.com
- Meta Muse AI may throw gas on the labor displacement problem — but it's not all doom and gloom finance.yahoo.com