The discussion begins with a critical look at proposed AI regulations, including a FINRA-like body for testing frontier models and a White House policy to limit US releases to the level of China's best open model, which the panel finds impractical. The focus then shifts to Mira Murati's 975-billion-parameter open model, Inkling, and the strategic value of reinforcement fine-tuning for customizing AI in sensitive industries. Ramin Hasani challenges the feasibility of recursive self-improvement in current fixed-weight models, predicting a future where models surpass human understanding. The conversation explores Liquid AI's worm-inspired neural networks, which pack more information into smaller units and are already deployed in Mercedes-Benz cars. Finally, the panel discusses Palmer Luckey's critique of the patent system, AI outperforming doctors on medical benchmarks, and a breakthrough in reversing glycation damage, a previously permanent aging process.