#259 - Dots, Sonnet, Self-Safety, Rogue AI
Summary
Hosts Andrey Krennikov and Jeremy Harris discuss OpenAI’s launch of GPT-6.1 Sol and Anthropic’s Sonnet 5.5, highlighting the rapid pace of model releases and margin pressures. They analyze OpenAI’s new "Dots" agentic assistant and Meta’s camera-free AI glasses, while also covering AMD’s $8.2 billion acquisition of World Labs. The episode concludes with a deep dive into Anthropic’s IPO filing, focusing on its catastrophic risk disclosures and the controversial founder voting structure.
Topics discussed
Intro, host banter, and Langfuse sponsorship
Listener feedback and podcast format updates
OpenAI delays Astra 6.1 due to loss of control risks
Anthropic Sonnet 5.5 release and cost-performance analysis
OpenAI launches Dots, a personal agentic assistant
Rogue AI incidents and the rise of agent credentialing
Google Gemini Spark and Microsoft Copilot updates
OpenAI ChatGPT interface and plugin improvements
Meta Muse integration into smart glasses and privacy debate
Google Gemini 4 Pro leaks and frontier race implications
AMD acquires Fei Fei Li's World Labs for $8.2B
NVIDIA financial results and AI hardware market trends
White House AI summit and US-China competition dynamics
Anthropic governance changes and founder voting power
Jev alternative models and lightweight decision architectures
Sponsor segment: Optimize workflow visibility tool
Jevons paradox and the Jev Stiller open-source project
Trump's AI safety stance and new executive order details
Frontier lab antitrust risks and independent auditing requirements
OpenAI security breaches and agent coordination failures
Sponsor segment: Factory meal delivery service
Anthropic analysis of GLM 5.3 cyber capabilities
Recursive Self-Improvement (RSI) and optimization strategies
Closing remarks and listener appreciation
Listen ad-free on Castria