AI News: Opus 5.5, GPT-6 Sol, Jev, Muse and More!
This week in AI saw major announcements from Meta Connect (Muse agent, new AR/VR glasses), three new language models (Claude Opus 5.5, GPT-6 Soul, Grok 4.7), significant buzz around Typesafe AI's Jev decision model, and new features from YouTube, Microsoft, Google, and Spotify. Claude Opus 5.5 emerged as the new state-of-the-art model while Jev introduced a fundamentally different approach to AI outputs focused on decision-making rather than text generation.
Summary
The week opened with Meta Connect, where the host attended in person and witnessed multiple announcements. Meta introduced Muse, an AI agent comparable to OpenAI's agents but with easier setup and no complicated configuration requirements. Muse features voice control, integration with Meta Ray-Ban glasses, computer use capabilities on Mac, expanded connectors including Granola, an email address for forwarding tasks, and an upcoming Muse Charm keychain device for holiday release. Meta also unveiled new glasses including Ray-Ban Meta audio (camera-free alternative to address privacy concerns) and their most impressive announcement: lightweight VR glasses at 100 grams with Apple Vision Pro-level quality, shipping spring 2027 for $1,299. These glasses feature full-body holograms for calling and leverage a separate battery puck for computing power. Additionally, Meta announced Horizon Create and Horizon Studio, AI-powered platforms allowing creators to describe 2D/3D games that are automatically built with progression systems and multiplayer functionality.
OpenAI released three major updates: GPT-6 Soul and GPT-6 Luna (price-cut models offering 50% cost reduction while maintaining strong performance), and general availability of GPT Live 1 API enabling voice assistants to work while agents operate in the background. Anthropic launched Claude Opus 5.5, now the state-of-the-art model across multiple benchmarks including agentic coding, knowledge work, and computer use. Despite a 20% price reduction per token, Opus 5.5's token inefficiency means actual cost per task remains relatively high at $6 compared to GPT-6 Soul at $1.60. The model demonstrated extraordinary capabilities, generating a near-complete Mega Bonk game recreation over 20 hours and creating sophisticated animations using JavaScript and Blender integration.
SpaceX AI released Grok 4.7, positioned as an inexpensive option but ranking lower than both Opus 5.5 and GPT-6 Soul across benchmarks and showing weaker performance on creative tasks like game development.
Typesafe AI's Jev model generated significant buzz despite releasing last week. Unlike traditional LLMs that generate text, Jev outputs structured decisions (choices, scores, or boolean values) with confidence scores. This fundamentally different approach makes it extremely cheap (4 cents per million input tokens, free output) and fast, completing 27 questions in 11 seconds versus GPT-5.6 Terra's 9 seconds with higher cost. Use cases include smart file organization, email prioritization, comment moderation, malicious URL detection, and even video game playing based on frame-by-frame decisions. New signups are currently closed due to overwhelming demand.
YouTube announced custom feeds using AI, Ask YouTube for conversational search, auto-dubbing for live streams, AI-assisted thumbnail creation, new editing tools, likeness detection, A/B testing for intros and thumbnails, and Ask Music for conversational discovery across podcasts and 300 million songs.
Microsoft unveiled a new Copilot interface with unified home, code, and autopilot modes, integrating chat, deeper thinking, model selection (GPT or Claude), code environments, and automation capabilities across Microsoft Office suite.
Google released Gemini 3.8 Live with real-time video avatar generation paired with speech, text-to-speech with over 1,000 voice options, integrated Omni into Google Vids for dynamic animations, and announced Project Suncatcher to test TPU space deployment with SpaceX.
Spotify launched taste profile customization using AI to help users dial in their music preferences.
Key Insights
- Claude Opus 5.5 achieved state-of-the-art performance across multiple benchmarks (agentic coding, knowledge work, computer use) but remains expensive per task at $6 despite 20% per-token price reduction, because the model uses significantly more output tokens than competitors, making token-per-task cost higher than the per-token pricing suggests.
- Jev represents a fundamentally different AI model architecture that outputs structured decisions (choices, scores, booleans) with confidence scores instead of generated text, making it drastically cheaper at 4 cents per million input tokens and free output tokens, while completing complex decision tasks in seconds.
- Meta's new lightweight VR glasses weigh only 100 grams with Apple Vision Pro-level display quality achieved by offloading all computing to a separate battery puck connected via cord, rather than embedding processing in the glasses themselves.
- GPT-6 Soul costs half as much as its predecessor (input dropped from $4 to $2 per million tokens, output from $20 to $10) while maintaining nearly equivalent performance to GPT-6 Astra on most benchmarks, positioning it as the cost-effective alternative for users willing to sacrifice 10% performance.
- Claude Opus 5.5 demonstrated exceptional capability at animation generation using JavaScript and Blender integration, enabling models to not just generate visual designs but take control of external tools to produce sophisticated animations that previously required specialized software expertise.
Topics
Transcript
[0:00] This is another one of those weeks that's been absolutely insane. I just got back from Meta Connect where there was a ton of announcements, but we also got not one, not two, but three new models. The made on YouTube event was also this week where they announced a whole bunch of new AI features for YouTube. And for a couple days, the whole AI world went crazy about something called Jev. So yeah, we got a lot to talk about. I don't want to waste any more of your time. Let's get straight to it. Let's start with Meta Connect because well, I literally just got back from this event and it's the [0:31] hottest thing on…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Matt Wolfe
5 Ways To Use ChatGPT's New Features
The transcript demonstrates five practical applications of ChatGPT's new features, including computer use for software automation, advanced image editing, cloud browser integration for account access, voice-powered live agents, and persistent session agents for long-term user interaction tracking.
My BIGGEST AI Pet Peeve...
The speaker criticizes companies for announcing features before they're ready, using Claude's Co-work announcement as an example. They argue that companies should only announce features once they're fully available, not during rollout phases.
Opus 5.5 Is Crazy Good and GPT-6-Sol Launched Too
Anthropic released Claude Opus 5.5, a new state-of-the-art model that outperforms their previous best model (Fable 5.1) while being 40% cheaper and faster. OpenAI simultaneously launched GPT-6 Soul and Luna, which are cheaper and faster but less capable than their flagship Astra model, making Opus 5.5 the more impressive release of the day.
This Robot Looks DEMONIC... On Purpose?
A California company created Three Halves, a deliberately creepy-looking four-legged robot designed to perform dangerous tasks in hazardous environments like wildfires and collapsed buildings. Its anthropomorphic torso allows it to use human-designed tools, while its design features—including locked joints, stable stance, and horn-mounted cameras—prioritize safety and functionality over aesthetics.
He Built The Ultimate Spy Tool (Free and Open-Source)
Belaval Sadu created God's Eye View, a free open-source platform that aggregates real public data sources (flights, satellites, vessels, CCTV, traffic) into an interactive globe interface to provide civilian intelligence and situational awareness. The tool uses AI agents and multiple APIs to make geospatial data accessible to anyone, with applications ranging from content creation to journalism and disaster reconstruction.