AI does not yet matter to most people. Core daily life, food, family, transport, entertainment, runs almost entirely without it. The author, Nathan Lambert, returned from a weeks-long wedding in New England and used AI exactly twice: for search and to generate a seating chart. That anecdote is not a limitation of Lambert specifically. It is a precise measurement of where the technology sits in 2025 for the median American.
The industrial revolution comparison that AI optimists reach for actually undermines their case. The First Industrial Revolution delivered cheaper clothing and cookware. The Second delivered indoor plumbing, bicycles, preserved food, and electric light. Physical, immediate, impossible to ignore. Current AI's most plausible wins, solving the Navier-Stokes problem, accelerating rare disease therapeutics, compounding scientific infrastructure over decades, are benefits so indirect that no ordinary person will know whom to thank. Lambert introduces Engels' Pause here, the 1790 to 1840 period when British GDP expanded rapidly while working-class wages flatlined, and the parallel is uncomfortable: AI is already fundamental to knowledge work, roughly half the U.S. economy, while doing little for the other half. Tech sector headcount is shrinking even as output explodes.
The argument Lambert is building is not that AI will fail. It is that the timeline is misread and the political consequences of that misread are arriving now. If the people who do not benefit from AI identify it correctly as a tool for the elite, and they can, the backlash will constrain development before the compounding effects ever reach them. Read the full piece for where Lambert thinks the breaking point sits and what the next 18 months of agent deployment will test.
[READ ORIGINAL →]