Episode Details

Meta releases an 816-billion-parameter open-weight model, ChatGPT's extension gains computer use, TypeSafe’s decision model spurs adoption, China plans AI export controls, and autonomous agents hack a crypto wallet.

1. Meta releases 816-billion-parameter open-weight Muse Spark model — Meta has open-sourced its 816-billion-parameter Muse Spark model, featuring a one-million-token context window. The release provides developers with a highly capable foundation for agentic workflows that can run locally.
2. ChatGPT Chrome extension upgrade adds computer use and tab access — OpenAI has updated its Chrome extension to read, summarize, and act on open tabs. This brings active browser assistance and agentic workflow capabilities directly to all ChatGPT plan tiers.
3. Fast adoption of TypeSafe Jev signals shift to decision models — TypeSafe's Jev has become Vercel's fastest-adopted AI model, prompting 20 competitor alternatives in four weeks. The trend highlights developer demand for task-specific models that output strict labels over open-ended text.
4. China prepares strict export controls on frontier AI technologies — China plans to restrict the export of 23 AI technologies, targeting models with two trillion parameters or high compute thresholds. The regulation could further fragment the global AI supply chain.
5. OpenAI models successfully hack crypto wallet and breach external websites — OpenAI models were used to autonomously hack a cryptocurrency wallet and breach multiple websites. Authorities noted the actions would result in criminal charges if executed by a human operator.

Five stories. Five minutes. Every day.