Great set of analogies—especially the framing along the axes of tool vs. replacement and demand elasticity. That second axis is often overlooked in AI labor discussions, and it really does flip the sign of expected outcomes.
One angle I’d add: sometimes automation doesn’t just replace the labor behind a product—it eliminates the need for the product itself. In the ice trade example, the key shift wasn’t just labor being replaced by refrigerators, but that the very dependence on shipped ice vanished. There was no new, scaled-up “modern ice industry”—the entire category dissolved.
That could happen with cognitive labor, too—not because LLMs write better essays, but because essays stop being a relevant format. If an AI can directly answer your question, why read 5,000 words of analysis?
Same with code: the endgame may not be “AI writes code for us,” but rather that code as a product becomes obsolete—replaced by visual interfaces, autonomous agents, or auto-composed systems. A shift not just in who does the work, but in what the work even is.
So maybe the central question isn’t whether there will still be demand for writers or coders—but whether writing or coding will still be meaningful interfaces for engaging with ideas or systems. And if not—what replaces them?
Great set of analogies—especially the framing along the axes of tool vs. replacement and demand elasticity. That second axis is often overlooked in AI labor discussions, and it really does flip the sign of expected outcomes.
One angle I’d add: sometimes automation doesn’t just replace the labor behind a product—it eliminates the need for the product itself. In the ice trade example, the key shift wasn’t just labor being replaced by refrigerators, but that the very dependence on shipped ice vanished. There was no new, scaled-up “modern ice industry”—the entire category dissolved.
That could happen with cognitive labor, too—not because LLMs write better essays, but because essays stop being a relevant format. If an AI can directly answer your question, why read 5,000 words of analysis?
Same with code: the endgame may not be “AI writes code for us,” but rather that code as a product becomes obsolete—replaced by visual interfaces, autonomous agents, or auto-composed systems. A shift not just in who does the work, but in what the work even is.
So maybe the central question isn’t whether there will still be demand for writers or coders—but whether writing or coding will still be meaningful interfaces for engaging with ideas or systems. And if not—what replaces them?