A post outlines four stages for humanoid robotics, each dependent on progress in the previous stage: building strong hardware, pairing it with a whole-body, AI-first architecture, scaling intelligence, and scaling manufacturing while integrating robots into the economy. The author argues that the first two stages require deep engineering and cannot be brute-forced with capital. For intelligence, the author expects data and compute to become the main bottlenecks, ultimately requiring far more of both than large language models. Capital could accelerate intelligence and manufacturing, the post argues, but doing them correctly would require tens—and eventually hundreds—of billions of dollars.