Matt Wolfe

This weekly AI news roundup covers Claude Mythos, an unreleased AI model from Anthropic that's so powerful at finding software vulnerabilities they won't release it publicly. Instead, they're giving limited access to major tech companies through Project Glass Wing to help patch vulnerabilities before similar models become widely available.

A company called RC released an open-source AI model called Trinity Large Thinking that performs comparably to major models like Claude Opus in benchmarks. The speaker struggles to find meaningful ways to test new AI models for everyday use cases since most models already handle typical business and personal tasks adequately.

OpenAI has made history with a record-breaking $122 billion funding round at an $852 billion valuation, making it the fastest growing company ever with $2 billion monthly revenue. Despite speculation about conflicts, Microsoft participated in this funding round, while OpenAI's Sora product was shut down after losing $1 million daily.

OpenAI has acquired TBPN (Tech Business Production Network), described as the 'ESPN of tech and business,' which surprised the speaker given OpenAI's stated focus on avoiding distractions. The speaker questions how this acquisition will affect the show's editorial independence.

Anthropic's Claude source code was leaked across GitHub, revealing details about how the AI system handles memory and context storage. The leak showed that Claude uses an indexing system that stores location references rather than raw data, and searches through transcripts rather than fully loading them into context.

The biggest AI news this week was Anthropic's Claude Code source code being leaked, revealing a sophisticated memory architecture and an unreleased autonomous agent called Chyros. OpenAI raised a record $122 billion and announced plans for a unified AI super app, while Microsoft released MAI Transcribe 1, a new speech recognition model.

A benchmark report from Catata reveals a significant 25 percentage point accuracy gap in how AI models access data through different MCP (Model Context Protocol) implementations. While some approaches achieve only 65-75% accuracy by translating prompts directly to API calls, Catata's standardized relational interface with semantic context achieves 98.5% accuracy.

The video covers extensive AI news from the week, primarily focusing on Anthropic's prolific release schedule of 74 features in 52 days, including computer control capabilities and project management tools. Google released Gemini 3.1 Flash Live with multimodal conversations, while OpenAI surprisingly shut down Sora and other side projects to focus on core chat and coding capabilities.

Bernie Sanders and AOC introduced a bill to pause data center construction until AI safeguards are established, citing rising electricity costs for residents. The creator argues this could paradoxically strengthen big tech companies by constraining compute supply while demand remains high, ultimately hurting smaller businesses and individual users.

OpenAI is shutting down Sora, its AI video generation app and API, as part of a strategic shift to focus on productivity tools for businesses and coding. The company is redirecting computing resources away from what they consider 'side quests' toward their core ChatGPT platform and a new model codenamed 'Spud'.

This week's AI news covers major releases including MidJourney V8 (which received mixed reviews), Microsoft's MAI Image 2, and Google's new vibe design and coding tools. Nvidia's GTC conference dominated attention with announcements including Nemo Claw for secure OpenClaw deployment and projections of $1 trillion in GPU sales through 2027.

The video demonstrates how to run open-source AI models locally on an iPhone using a free app called Locally AI, without requiring an internet connection. The presenter showcases the newly released Qwen 3.5 model in various sizes and tests its capabilities for brainstorming, image recognition, and parenting advice. The key appeal is privacy and offline functionality, with performance comparable to state-of-the-art models from about two years ago.

A YouTuber and content creator shares the AI tools he uses daily in 2026, ranked from most to least frequent use. The tools span search, writing, coding, meeting notes, voice synthesis, and news aggregation. He emphasizes practical productivity gains over theoretical AI potential.

The video is a comprehensive tutorial and review of 'Claudebot,' an open-source AI agent that runs on a local machine or VPS, can take autonomous actions on a computer, and communicates through messaging apps like Slack. The creator walks through setting up Claudebot on a free AWS EC2 instance, connecting it to Slack, and demonstrates use cases like daily news aggregation, app building, and animation generation. He concludes with serious security warnings about prompt injection and the risks of giving an AI agent broad system access.

A creator demonstrates building a fully playable video game called "Library Survivors" in a single day using AI tools like ChatGPT, Claude Code, Leonardo, and Suno AI, without prior coding experience. The process involves conceptualizing the game, using AI to flesh out mechanics, generating code, iteratively debugging and refining gameplay, creating graphics, and adding sound effects.