Last Week in AI #251 - Mythos Back, Sonnet 5, Etched, LongCat
Summary
Hosts Andrey Krenkov and Jeremy Harris discuss the US government lifting restrictions on Anthropic’s models while delaying OpenAI’s GPT 5.6, analyzing the geopolitical implications for AI safety and competition. They cover Anthropic’s new Claude Sonnet 5, Google’s Notebook LM updates, and infrastructure startup Etched’s rapid chip development. The episode also touches on Baidu’s AI chip IPO, Agility Robotics’ SPAC listing, and DeepSeek’s hiring expansion.
Topics discussed
Introduction and episode overview
Anthropic's jailbreak response and US government pressure
Anthropic Sonnet 5 launch, pricing, and benchmarks
Nano Banana AI art and video tool updates
Groq's custom AI chip and manufacturing progress
Kunlunxin AI chip IPO and valuation
Agility Robotics SPAC merger and humanoid robot hype
DeepSeek hiring spree and organizational shift
Meituan's DeepSeek-V3 model and Chinese AI efficiency
New AI agent benchmarks: GUI vs Terminal use
Taiwan probe into Nvidia chip smuggling to China
Meta's synthetic data and RL research papers
Neon Studios acquires OpenAI biopic 'Artificial'
Tidal AI music labeling and episode outro
Listen ad-free on Castria