In this inaugural "Office Hours" livestream, Anjney Midha and Mike Abbott field rapid-fire student questions covering the AI compute and infrastructure landscape. They discuss how scaling laws continue to hold in verifiable domains like coding, materials science, and robotics, while creative writing remains stubbornly hard—an area Anjney would love to see students attack with carefully curated data sets. On infrastructure, they argue that meaningful frontier research now requires clusters in the 4K-16K chip range, that the training/inference distinction is dissolving into one fungible compute pool, and that GPU price corrections will come through standardization and a "universal kernel" abstraction. The conversation closes with leadership lessons from Mike's time at Apple, GM, and Twitter—culture and fiscal discipline are one-way doors you cannot recover once broken—and an extended metaphor about company-building as a road trip where mission alignment determines whether your passengers tolerate the scenic detours. They wrap with a peek at AMP's "grid" scheduling system, which targets Google-level node utilization (95%+) across a multi-cloud portfolio.