#245 - TML-Interaction, Claude For Legal, Sam Altman on Stand
Summary
Hosts Andrei Krenkov and Jeremy Harris discuss OpenAI’s new real-time voice API and Thinking Machines’ low-latency conversational model, analyzing the trade-offs between reasoning depth and speed. They examine Anthropic’s “Claude for Legal” launch, debating whether foundation models will cannibalize specialized AI startups like Harvey. The episode also covers Meta’s integration of AI into Threads, highlighting the shift toward embedded, agentic interactions in social platforms.
Topics discussed
Intro, host banter, and sponsor read
OpenAI GPT-5 and real-time voice features
Deep dive into GPT-5 architecture and latency
Anthropic's new tools and legal AI competition
Meta's AI integration and Google's agentic updates
OpenAI board drama and Sam Altman controversy
Jensen Huang's trip to China and geopolitics
AWS Anthropic deal and Chinese gray market APIs
AI in drug discovery and independent model auditing
OpenAI security partnerships and alignment research
Jack Clark on automated AI R&D and takeoff
ChatGPT safety features and deceptive chain of thought
Critique of Anthropic's risk assessment surveys
Celebrity AI registry and AI capability benchmarks
Listen ad-free on Castria