AI News: Claude's Massive Leap & Siri Gets Good!?
This AI news roundup covers the release of Claude Fable 5 (a Mythos-tier model from Anthropic) and its controversial safety restrictions, Apple's WWDC AI announcements including a major Siri overhaul, and updates from Google including NotebookLM upgrades and real-time translation. Additional rapid-fire items include OpenAI and SpaceX IPO filings, ChatGPT email sending, and a teased Midjourney hardware device.
Summary
The video opens with a breakdown of Anthropic's June 9th announcement of two models: Claude Fable 5 and Claude Mythos 5. The host clarifies that while Fable 5 is described as a 'Mythos class' model sitting above Opus in Anthropic's tier hierarchy, it is not the same as the Mythos 5 model, which is restricted to a small group of cyber defenders and infrastructure providers. Fable 5 includes heavy safety restrictions around cybersecurity, biology, chemistry, and notably, frontier LLM development — where the model silently downgrades its responses without informing the user. This sparked significant backlash from figures like the Hugging Face CEO and others, leading Anthropic to partially backtrack within two hours of release, committing to transparency about when safeguards are triggered. Pricing is set at $10 per million input tokens and $50 per million output tokens — double that of Opus 4.8. The host demonstrates impressive one-shot coding capabilities, building a Mega Bonk game clone and a YouTube clone with a recommendation algorithm using minimal prompts, as well as a desktop B-roll generator tool built in just two prompts.
The video then covers Apple's WWDC, which was heavily focused on AI. Apple announced a deep collaboration with Google to use Gemini models alongside their own foundation models for Apple Intelligence experiences. The new Siri is positioned as a major upgrade — more conversational, context-aware across device (photos, calendars, messages), and capable of visual intelligence via the camera. New features include a dedicated Siri app with cross-device memory, Siri coming to Spotlight on Mac, Apple Watch, and Apple Vision Pro, AI-powered shortcut creation via natural language description, and AI image generation and editing. The host also demonstrates the spatial reframing feature, available now in the iOS 27 beta, which uses generative AI to fill in areas outside a photo's original frame. The host notes these features won't initially be available in the EU.
Google's announcements include a NotebookLM overhaul powered by Gemini 3.5, now equipped with a secure cloud computer, over 100 curated skills, and new output formats. Gemini 3.5 Live Translate was showcased as a near real-time translation tool coming to Google Meet, Google Translate, and available now in AI Studio. Google also introduced Diffusion Gemma, a text generation model using diffusion-based architecture (similar to image generation models) that generates entire 256-token paragraphs simultaneously, dramatically improving speed on local hardware.
In rapid-fire coverage, the host summarizes articles by Dario Amodei (calling for FAA-style AI regulation, UBI, international AI coalition) and Sam Altman with Jacob Pachocki (predicting AI-driven research by March 2028, emphasizing broad distribution of AI benefits). Other items include ChatGPT's new in-app email sending feature, OpenAI's IPO filing, SpaceX's record-breaking $75 billion IPO at a $1.77 trillion valuation, Coinbase opening its platform to AI agents, and a teased hardware launch from Midjourney founder David Holtz.
Key Insights
- Anthropic's Fable 5 silently downgrades its responses when users ask about frontier LLM development — the model still responds but gives worse answers without informing the user, a practice Anthropic's own documentation confirmed and which sparked widespread criticism from the AI community.
- Apple announced a deep collaboration with Google to power Apple Intelligence experiences using Gemini models alongside Apple's own foundation models, marking a significant shift in Apple's AI strategy toward third-party partnerships.
- Google's Diffusion Gemma generates entire 256-token paragraphs simultaneously rather than word-by-word, allowing it to fully utilize local GPU/TPU hardware — addressing a key inefficiency of traditional autoregressive models when run on personal devices.
- Dario Amodei argues for an FAA-style regulatory body to review AI models before public release, international coalitions to prevent any single country from dominating AI, and UBI combined with higher taxes on AI companies to distribute economic gains broadly.
- Sam Altman and Jacob Pachocki predict that by March 2028, a significant fraction of OpenAI's research may be conducted by AI systems working in tandem with human researchers, framing this as the beginning of a 'third phase' of OpenAI where the economy shapes itself around AI.
Topics
Transcript
[0:00] Here's the AI news you probably missed this week. Starting with the brand new model that has everybody talking right now, Claude Fable 5. Now, this is a model that I did an entire breakdown video of, where I explain exactly what this model is, what people love about it, what people hate about it, some misconceptions, and even test it a little bit. You can find it over on my YouTube channel. It's called Why Everyone Is Freaking Out Mythos. At least for right now. I split test titles and thumbnails constantly. But in case this is like the only video you watch [0:32] each week to stay up to date with the news, I'll give you the…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Matt Wolfe
AI News: GPT-5.6 and the new Super App are a Massive Leap!
OpenAI released GPT-5.6, a major leap forward in AI capabilities, alongside the new unified ChatGPT work app integrating code, browsing, and agent features. The week also saw significant model releases from xAI (Grock 4.5), Meta (Llama Spark 1.1), and research from Anthropic on AI reasoning patterns, establishing a new competitive landscape in AI development.
I Built A Monetizable Business With AI
The creator built a monetizable finance dashboard in one day using Hyper Agent from Airtable, which deploys a team of AI agents that continuously monitor news, track funding rounds and IPOs, analyze market sentiment, and generate daily briefings without manual intervention. The system demonstrates how specialized AI agents can work together autonomously to create a functional business product.
The ONLY AI Benchmark You Need!
A developer created "Buccy Bench," a humorous yet functional AI benchmark that tasks different language models with drawing Gary Busey as SVG code rather than images. The benchmark tracks model evolution over time while measuring performance metrics like cost, tokens, and execution speed.
GLM-5.2 - The Open Model That's As Good As Opus!
A comprehensive review of GLM-5.2, an open-weight Chinese AI model with a 1 million token context window, demonstrating its capabilities for coding, document analysis, and agentic workflows at significantly lower costs than frontier models like Claude Opus and GPT-4.5. The speaker tests various use cases including website building, Chrome extensions, game development, and data organization, concluding it's valuable for long, code-heavy, token-expensive tasks despite not universally outperforming closed-source alternatives.
Don't Fall For This AI Trap
The speaker emphasizes that power users distinguish themselves by knowing what NOT to automate with AI, rather than automating everything. They argue that AI works best for clear, straightforward tasks but struggles with nuanced, artistic work requiring consistency—using their failed YouTube thumbnail automation as an example.