Now Playing
Nobel Laureate Leaves DeepMind, GLM-5.2 Shocks Silicon Valley & Export Controls Bite
Audio is available on Spreaker — see link below.
John Jumper's move from DeepMind to Anthropic signals a talent shift at the frontier, while China's GLM-5.2 matches GPT-5.5 on coding benchmarks and U.S. export controls accidentally shut down Anthropic's own models. Five stories reshaping the AI competitive landscape in under 15 minutes.
Audio is available on Spreaker — see link below.
John Jumper, the Nobel laureate who co-led AlphaFold at DeepMind, just walked out the door. After nine years, he's joining Anthropic.
Meanwhile, out of China, something significant landed this week. Zhipu AI released GLM-5.2, a seven hundred forty-four billion parameter open-weight model that's generating a level of attention in Silicon Valley not seen since DeepSeek.
Here's where the regulatory picture gets genuinely complicated. A U.S. government export control directive this week blocked access to Anthropic's Fable 5 and Mythos 5 models for foreign nationals.
On the acquisition side, SpaceX announced an all-stock deal to acquire AI coding startup Cursor, valued at sixty billion dollars. The deal is expected to close in the third quarter.
OpenAI isn't standing still. GPT-5.6 variants are expected next week, featuring a one-and-a-half million token context window, improved coding performance, and pricing designed specifically to undercut Anthropic.
Two things to track closely from here. First, whether DeepMind's talent loss stays isolated or accelerates.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.