Meta’s Muse Wonder - Fable Class Model
Read the Portuguese translation →
Author: Prompt Engineering · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: Meta’s Muse Wonder - Fable Class Model · Channel: Prompt Engineering · 01:01 · Watch on YouTube
Topics: AI agents, AI evaluation and benchmarks, AI models
Summary
Section titled “Summary”Meta introduced Muse Wonder as a model focused on “personal super intelligence” for everyday tasks. Its two highlighted features are active memory, which carries relevant information across conversations, and active focus, which works through long tasks and checks in before important actions. A demo shows it using a browser and other apps to create a running plan based on a user’s preferences. Meta says Muse Wonder is available through its API, while a larger version is part of Meta’s code subscription.
Key points
Section titled “Key points”- Meta describes the goal as personal super intelligence for everyday use.
- Meta created an internal benchmark called the personal super intelligence index.
- Active memory carries relevant information about the user across conversations.
- Active focus continues working on long tasks and tries another approach if it gets stuck.
- The model checks in with the user before taking important actions.
- A demo shows it using a browser and other apps to create a personalized running plan.
- Muse Wonder is set to power Meta’s personal agent, starting with paid subscribers.
- Meta says a larger version leads long-horizon coding benchmarks, including Deep Sea.
- Muse Wonder is available through Meta’s API; the full version is available through Meta’s code subscription.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | Personal super intelligence | Meta introduces Muse Wonder and its focus on useful everyday assistance, along with a new internal benchmark. |
| 00:30 | Active focus and personal assistance | The video describes active focus, a personalized running-plan demo, and Muse Wonder’s role in powering Meta’s personal agent. |
| 01:01 | Availability and coding benchmarks | Meta says a larger version leads long-horizon coding benchmarks and outlines how Muse Wonder and the full version can be accessed. |
Quotes
Section titled “Quotes”Meta just released Muse wonder — 00:00
useful for people in their everyday lives — 00:00
The second one is active focus — 00:30
build a running plan across your preferences. — 00:30
Muse wonder is available today through the Meta’s API — 01:01
Concepts and entities
Section titled “Concepts and entities”Meta, Muse Wonder, personal super intelligence, Personal Super Intelligence Index, active memory, active focus, Muse, Deep Sea
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 Meta just released Muse wonder and the whole launch is built around what Meta calls personal super intelligence. Which means models that are useful for people in their everyday lives rather than just yet another coding agent. Meta even created a new internal benchmark for this called personal super intelligence index. Now Meta is focusing on two features. The first one is active memory which means the model carries what matters about you across conversations. So you don’t have to
00:30 re-explain yourself every time. The second one is active focus which keeps the model working on long tasks trying a different approach when it gets stuck and checking in with you before it does anything important. So in one of their demos wonder works across a browser and other apps to build a running plan across your preferences. This is also the model that is going to power Muse which is Meta’s personal agent starting with paid subscribers. Now it can still code and according to
01:01 Meta the bigger version called Muse wonderful leads long horizon coding benchmarks like Deep Sea. Muse wonder is available today through the Meta’s API while the full version is only inside Meta’s code subscription. So it’s going to be really interesting to see whether this personal focus actually shows up when people start using it.