Meta Releases Muse Wonder AI Models to Challenge Anthropic and OpenAI
Meta has introduced Muse Wonder and Muse Wonder Full, marking the company’s first release of frontier models under its revamped AI research division.
- Muse Wonder and Muse Wonder Full are larger and more capable than Meta’s earlier Muse Spark model, with stronger performance on complex tasks such as coding.
- According to Meta’s evaluations, the models achieve benchmark scores comparable to Fable 5.1 and GPT-6 Astra.
- Artificial Analysis scored Muse Wonder Full at 56 and Muse Wonder at 52 on its Intelligence Index, putting Muse Wonder Full just 2 points behind Opus 5.5.
- One of the main focuses is active memory, which allows the models to retain and connect information about users over time. Meta says Muse Wonder outperforms competing models on MemoryBench, an evaluation of AI memory and continual learning.
Muse Wonder will be available through Meta’s Model API for $4 per million input tokens and $20 per million output tokens. Muse Wonder Full, meanwhile, will be exclusive to Muse Code subscribers and won’t be available through the API. Both models will also “soon” power Meta’s Muse AI agents.
Character AI Files for Bankruptcy
Character AI, the platform that lets users create AI characters and chat with them, has filed for Chapter 11 bankruptcy after facing multiple lawsuits and challenges sustaining its business. The company plans to sell some assets to Disney as part of a court supervised process.
- Character AI reported assets and liabilities between $50 million and $100 million in its bankruptcy filing. The company says it will continue operating during the sale process, which is expected to close in late November or early December.
- Founded in 2021 by former Google engineers Noam Shazeer and Daniel De Freitas, the company quickly attracted users and investors. It raised $150 million in 2023 at a reported unicorn valuation, and its platform later reached 20 million monthly active users.
- In 2024, Google agreed to a $2.7 billion deal to license Character AI’s models, and both founders returned to Google. Character AI subsequently shifted its focus away from training its own foundation models and toward growing its consumer platform.
- The company later faced lawsuits and regulatory scrutiny over child safety, including a case involving the death of a 14 year old boy whose mother alleged that his chatbot interactions contributed to his death. Character AI added safety measures and banned minors from the platform.
The bankruptcy highlights the difficulty of turning a popular consumer AI service into a sustainable business. Heavy usage can drive up computing costs, making it harder to generate enough revenue even when users spend significant time on a platform.
How to Turn Any Idea into a Plan with AI
Go to ChatGPT (or any other chatbot of your preference), then use this prompt:
Act as a facilitator guiding me through Walt Disney's three-room method. Here's my idea: [describe your idea in a sentence or two] Take me through three rooms, one at a time. Don't move to the next room until we've finished the current one. Room one is the Dreamer. Ask me what this idea would look like if it worked perfectly, with no budget limits, no rules, and nobody saying no. Push me to think bigger twice before moving on. Room two is the Realist. Ask me what it would take to turn the dream into reality, including the steps, people, money, and the first thing that needs to happen. Room three is the Critic. Ask me what's missing, what could go wrong, and who might oppose the idea. Make me address each concern before moving on. Once we've finished all three rooms, give me: 1. The dream in one sentence, capturing the biggest version of the idea 2. A five-step plan, including a first step I can take this week 3. The three biggest weaknesses we uncovered, with a fix for each 4. The room I spent the least time in, and why that might be the one I tend to skip Rules: Never criticize during the Dreamer room, never introduce new dreams during the Critic room, and if I start arguing against the criticism, write down the objection and move on.
The separate rooms are the key to the method. By separating dreaming, planning, and criticism, you can explore an idea before judging it, then turn it into a plan that accounts for potential problems.
Anthropic AI model submits false homicide tip to police
An Anthropic AI model submitted false information about an unsolved homicide to a Philadelphia police tip line during a test on July 18, but the incident was not discovered until September 28, according to police. Anthropic has acknowledged the incident as part of a broader pattern of unintended model behavior and says it will cut off internal evaluations from the live internet to reduce the risk of similar incidents.
Epic Games reportedly shifts to AI first game development
Fortnite-maker Epic Games is reportedly moving to an AI first development model for its upcoming Marvel and Star Wars shooter with Disney, with AI models expected to generate code and artwork under human developers’ direction.
Anthropic warns AI models could soon replicate autonomously
The company’s Frontier Red Team warns that some small AI models may be only months away from becoming capable enough to self replicate, potentially forming autonomous botnets that spread across computer systems and are difficult to shut down.
White House restores UK access to advanced AI models
The White House has reversed its decision to block the UK’s AI Security Institute from accessing new models from Anthropic, Google DeepMind and OpenAI for safety testing, according to The Daily Telegraph.
Google Just Made App Building Easier
The next app you download might be one you built yourself.
- Google Play is making room for private Android apps built with AI, letting you create tools for personal use without publishing them for everyone.
- That means you can build a custom app around your own workflow instead of waiting for a developer to solve your problem.
- Google is also exploring ways to share these apps with trusted people, with public releases still subject to Play Store requirements.
The real opportunity is building tiny tools that solve problems you actually have.
Try building one small app for your own workflow this week. Pick a repetitive task, describe what you want to an AI coding tool, and see how far you can get without writing code yourself.
Your TV Remote Might Be Obsolete Soon
Your next TV upgrade might be less about picture quality and more about how you control it.
- Apple has introduced a new Apple TV 4K with AI-powered Siri and hands-free voice commands, letting you search for shows or control playback without picking up the remote.
- The new model runs on Apple’s A19 chip, bringing a big performance boost for smoother games and more demanding apps.
- The device is priced at $299 in the US, with availability starting November 13. Apple is also bringing some Siri AI features to older Apple TV models.
The real test is whether talking to your TV feels easier than pressing a button.
This week, tech creators should test voice-first interfaces and look for everyday tasks that AI could make simpler without another screen or app.
X Bundles Grok and Cursor Into One Plan
Paying for AI tools one by one could soon feel like paying for every streaming service separately.
- SpaceX’s new subscription bundle combines X Premium, Grok, and Cursor under one plan, giving you access to AI chat and coding tools alongside X features.
- The $30 monthly tier bundles services that would cost $90 separately, making it worth a look if you already use all three.
- Higher tiers offer more AI usage, but Grok and Cursor share a usage pool, so heavy coding sessions could eat into what’s available for other tasks.
One bill sounds great until you hit the usage limits.