What you'll learn

• 1 (00:02) **China's Open-Source AI Resets the Race** - GLM5.2 from Z.ai matches U.S. frontier models on security benchmarks, reigniting the open vs. closed source debate.
• 2 (05:27) **Analyzing the Benchmark Chart** - Tyler explains the ELO-based chart from the Center for AI Standards and Innovation, which shows U.S. models ahead but with a contested gap.
• 3 (07:43) **Revisiting the Closed-Source Thesis** - John Ludig's 2024 prediction that open-source would lose the capex war is re-evaluated in light of China's strategy.
• 4 (12:30) **Tyler's Review of GLM5.2** - A deep dive into the model's performance, distillation concerns, and market fit.
• 5 (19:00) **Google Restricts Meta's Gemini Access** - Capacity constraints force Google to limit Meta's use of Gemini, disrupting internal AI projects.
• 6 (21:37) **Meta's Brain-Computer Interface Milestone** - Brain-to-Cord V2 can decode words and semantics from raw brain signals in real-time using a non-invasive device.
• 7 (23:53) **How Terry Semmel Fumbled Yahoo's Facebook Deal** - The story of Zuckerberg agreeing to sell for $1B in 2006, only for Semmel to cut the offer to $800M.

+ Full timestamped outline available in the app