Every major technology of the past century — electricity, semiconductors, the internet — made human intelligence more valuable by automating physical or narrowly defined cognitive tasks. Artificial intelligence breaks that pattern entirely. For the first time, we are developing a technology that could substitute for intelligence itself, and the economic implications of that distinction deserve serious attention from anyone allocating capital or building organizations today.
Stanford economist Charles Jones frames the challenge with admirable precision: the question is not whether AI will be transformative, but whether it will be transformative in a historically familiar way or in a categorically different one. His two-scenario framework is worth internalizing. In the accelerated growth scenario, AI raises the productivity of software engineers and AI researchers, who in turn build better models, which further accelerate research — a compounding loop that could eventually produce what Anthropic’s Dario Amodei describes as “a country of geniuses in a data center.” The empirical signals supporting this trajectory are already visible. Epoch AI estimates that effective compute used in AI training is rising tenfold annually. The benchmark measuring how long AI takes to complete complex software engineering tasks is halving every five to seven months. What required five hours of human effort eighteen months ago now takes nineteen minutes. 📅 POTENTIALLY STALE — this article reproduces Jones’s Jan 2026 framing. METR TH1.1 (Feb 2026) updated the frontier to Claude Opus 4.6 at 14.5 hours (not “five hours”). See Agentic AI Fundamentals and AI as a General Purpose Technology for current values.
The alternative scenario — “business as usual” — is equally plausible and equally important to hold in mind. General purpose technologies historically disappoint in the short run and surprise in the long run. Complementarity between tasks creates bottlenecks; automating ninety percent of a workflow may yield far less than ninety percent of the productivity gain if the remaining tasks are the binding constraints.
This is precisely where Jones’s “weak links” framework becomes strategically useful. Economies and organizations are systems of complementary tasks, and the value of automating any single node depends heavily on what surrounds it. For executives, this reframes the AI investment question: the return on automation is determined not by what you eliminate but by what you leave intact.
The honest conclusion is that the range of plausible futures remains extraordinarily wide. But sophisticated actors cannot afford to treat that uncertainty as permission to wait. The asymmetry of outcomes — between a world of compounding AI-driven growth and one of incremental productivity gains — demands that strategy be built to remain coherent across both.
Source: Raw/trigger-ai-and-our-economic-future.md