In this episode of The Hardware Podcast, Lucas and Luna explore how artificial intelligence is transforming the chip design process itself. They focus on a specific breakthrough: Google's use of reinforcement learning to automate floorplanning for Tensor Processing Units, a task that once took human engineers months. The conversation covers the scale of the problem—modern chips have billions of transistors—and why traditional EDA tools are hitting limits. They discuss the surprising 2020 Nature paper where an AI designed a chip floorplan in under six hours that matched or beat human results on key metrics, and how this has sparked a wave of similar efforts at NVIDIA, AMD, and Synopsys. Lucas explains the concrete numbers: reduced design cycles by weeks, improved wirelength by 5-10%, and lower power consumption. Luna pushes back on the hype, asking whether AI is replacing engineers or just changing their jobs. The episode ties into the broader hardware-software co-design trend and what it means for the pace of Moore's Law.