#212 - o3 pro, Cursor 1.0, ProRL, Midjourney Sued
Summary
This episode covers OpenAI’s O3 Pro release, Cursor’s 1.0 agentic coding features, and new models from Mistral, Eleven Labs, and ByteDance. It highlights Anthropic’s talent advantage over OpenAI and DeepMind, alongside OpenAI’s legal battle over chat logs. The hosts also analyze Huawei’s chip struggles against NVIDIA and discuss the shrinking entry-level tech job market due to AI automation.
Topics discussed
Sponsors: Box and Thrive Cosmetics
Intro, host banter, and episode structure
OpenAI O3 Pro release, pricing, and Cursor agents
Mistral's new reasoning models and open source strategy
Eleven Labs text-to-dialog API and emotional speech
Google Veo 3, ByteDance, and video generation risks
AI talent war: OpenAI, Anthropic, and DeepMind
OpenAI court order to preserve ChatGPT logs
Huawei chips, SMIC, and China's AI hardware race
Mistral Compute cloud and sovereign AI infrastructure
Pro-RL: New reinforcement learning techniques
Memory-aware scaling laws for LLMs
Positive vs negative reinforcement in reasoning models
AI automating scientific research and experiments
AI detecting alignment issues and in-context learning
Echo Leak vulnerability and Anthropic's national security
Disney vs MidJourney lawsuit and voice cloning rights
Listen ad-free on Castria