I Used to Hate Meta: Muse Wonder Changed My Mind

What you'll learn

• 1 (00:06) **Meta's Comeback: Muse Wonder Enters the Arena** - The host introduces the episode's thesis: Meta's new Muse Wonder model is a legitimate competitor to Claude Opus 5.5 and GPT-6 Astra, marking a major reversal from 18 months ago when Meta's models were lagging.
• 2 (01:32) **The Two Models: Muse Wonder and Muse Wonder Max** - The host breaks down the two variants of the new Meta model: the standard Muse Wonder (Opus 5.5 competitor at $4/$10 per million tokens) and the larger Muse Wonder Max (Fable 5.1 competitor, behind a $50/month subscription).
• 3 (02:17) **Benchmarking the Leap: From 31 to Second Place** - The host uses a tweet from Aaron Levy and the Artificial Analysis Intelligence Index to show Meta's model progression over 18 months, from scores of 31-34 to now placing second, just behind Claude Opus 5.5.
• 4 (03:27) **The Legal Edge: Doubling Second Place** - The host highlights a specific benchmark from Harvey Lab, which tests AI's ability to automate lawyer work. Muse Wonder scored 13.1, more than double the second-place model, Claude Fable 5.1 Max.
• 5 (04:40) **Coding Capabilities: Good, Not Great, But Unique** - The host discusses the model's coding performance, noting it's a major leap from its predecessor but not necessarily better than Claude Code or GPT-6 Astra Codex for building new apps.
• 6 (06:00) **Active Memory: Proactive and Intelligent Personalization** - The host explains Meta's "active memory" feature, which goes beyond simple memory by proactively pulling the most relevant context about the user into every conversation.
• 7 (07:58) **Active Focus: The Killer Feature for Non-Coders** - The host introduces "active focus," a feature where the AI automatically overcomes obstacles (e.g., paywalls, Cloudflare issues) by seeking alternative resources without asking the user for help.

+ Full timestamped outline available in the app