Even with physical AI, the worksite remains — author

The essay builds its case in two directions at once. One is capability: large language models first rippled through software development, and the author sees them spreading in order through writing, translation, clerical work, legal and accounting — all cognitive. The next step is a body, which is why the author points to NVIDIA's January 2025 CES keynote, where the company called physical AI AI's next frontier and announced Cosmos, a platform of world foundation models meant to speed up robot and self-driving development. The author's view is that the direction does not change even if the timing varies.

The other direction is human need. The essay separates 'can AI replace this' from 'will humans still need this', and answers the second: as long as people eat, wear clothes and need places to live, and as long as roads, bridges, tunnels, water and electricity, hospitals, government offices and schools are required, work that acts on physical things in physical places stays. What shifts is what the human does there. The essay sketches three stages — humans working, humans and machines collaborating, humans supervising machines — and applies them sector by sector, from construction to taxis to hospitals.

It also draws a line between two kinds of surviving worksite. One is functional, where substitution advances in steps. The other is relational — a hair salon, a bar, a sushi counter — where the draw is the person's charisma rather than the service itself, so replacing the function does not replace the place. The essay closes on motivation: even in an era of driverless taxis, someone may keep driving a private taxi not for income but because they want to, which is what the title means.