When will average people feel AI’s impact?

What it claims

The summary covers the thesis. Beyond it, the essay adds several specific claims and predictions:

• Knowledge work is about half the U.S. economy. For that half, AI is, or soon will be, "as fundamental as electricity." Lambert expects rapid improvement in agents over the next 18 months.
• Tech headcount will likely shrink while knowledge-work output explodes. Lambert endorses this view from Doug O'Laughlin (Fabricated Knowledge). Because AI's gains land in sectors that were already the most successful, it risks being branded as not a collective good.
• The worst case is the American nuclear power path. Lambert names this as the outcome to avoid: public backlash stunts AI's development.
• The industry faces two separable problems in what Lambert calls "the first half decade of a 50-year diffusion process": 1. Early benefits are too indirect. A patient credits the family doctor, not OpenAI, for a new cure, and few people care about AI solving a Navier–Stokes Millennium Problem. 2. The political backlash is inherited from Big Tech (Google, Meta) and is "only an AI story due to timing." Had AI's growth come decades later, Lambert argues, data centers would not be a central political issue.

Solving either would buy the industry time.

• The industry's own messaging deepens the backlash. "Proclamations of doom and mass unemployment" brand AI as negative or unsafe.
• The true adoption marker is deep integration, not chat. That means AI embedded in businesses and acting as personal assistants. It is "just starting to become viable" and will take far longer to adopt than ChatGPT. Lambert expects it to go from roughly 0% to 90%+ adoption within the lifetimes of younger people following AI today.
• Self-driving has developed on a trajectory "very independent" from LLMs. Lambert still expects robotics and self-driving to be folded into the LLM story, which he calls ironic given the effort spent arguing that LLMs differ from the previous decade of AI.

Limits

• This is an opinion essay with no data. The only figure, knowledge work as roughly half the U.S. economy, is unsourced.
• The 50-year and 18-month horizons are predictions, not derived estimates.
• The source for the Engels' pause concept appears only as a footnote marker; the footnote itself is not in the page text.
• The page ends by pointing to a Doug O'Laughlin piece. Its content is not included.
• The evidence is anecdotal: the author's own non-work AI use over a few weeks was search and making a wedding seating chart.
• Nothing here bears on choosing, adopting or building a specific model, dataset or tool.