The macro/micro distinction is Joel Mokyr's analytical framework for understanding how major technological transitions unfold across time. A macro-invention is radical and discontinuous — a device or technique that operates on principles sufficiently different from anything that preceded it that its potential applications cannot be deduced from its initial form. The steam engine was a macro-invention. The printing press, electrification, the transistor — each a discontinuity that opened possibility spaces requiring decades of subsequent exploration. A micro-invention, by contrast, is an incremental improvement that exploits the potential opened by a macro-invention. James Watt's separate condenser was a micro-invention. So were the thousands of subsequent refinements — higher-pressure boilers, better metallurgy, new applications from pumping to locomotion — that transformed Newcomen's crude pump into the engine of the Industrial Revolution.
The framework has three consistent features across every major technological transition. First, initial applications are conservative — the new capability applied to existing problems in existing domains. Newcomen's engine pumped water out of coal mines; the printing press initially reproduced Bibles and classical works. Second, the transformative applications emerge later, often decades later, from practitioners who grew up inside the new possibility space. The railroad came 118 years after Newcomen's engine. Third, the pace and direction of the cascade are determined not by the technology alone but by institutional infrastructure that supports experimentation, rewards creativity, and channels gains.
The AI transition shows every sign of following the same trajectory. The natural language interface that crossed the threshold in December 2025 is, by Mokyr's criteria, a macro-invention — a qualitative discontinuity, not merely an incremental improvement on prior AI systems. The applications visible in 2026 — coding assistants, writing tools, image generators — are conservative micro-inventions, the new capability applied to existing problems. The transformative micro-inventions remain in the future.
Mokyr identified four stages in the cascade: conservative application, creative adaptation, systematic exploitation, and combinatorial explosion. Most organizations in 2026 are in the first stage. The Trivandrum training described in The Orange Pill — engineers using AI to cross domain boundaries — represents creative adaptation. Systematic exploitation will require organizational redesign. Combinatorial explosion will combine AI with biotechnology, advanced manufacturing, and educational innovation to produce applications that no single technology could generate alone.
The historical discipline the framework imposes is temporal. The cascade from macro-invention to full exploitation took more than a century for the steam engine, roughly seventy years for electrification, approximately forty years for the personal computer. Each transition has been faster than the last, but none has been instantaneous. The limiting factor is not the technology but the institutional, organizational, and cultural adaptation required to exploit it.
Developed in The Lever of Riches (1990) and refined across Mokyr's subsequent career. The framework has become standard in economic history and is now being applied extensively to the AI transition.
Macro-inventions are discontinuities. They cannot be predicted from prior trajectory; they open new possibility spaces that require decades to explore.
Micro-inventions explore the space. The incremental refinements — efficiency improvements, new applications, new combinations — that transform a crude initial instantiation into a mature technology.
Four-stage cascade. Conservative application, creative adaptation, systematic exploitation, combinatorial explosion — each stage takes years or decades.
Institutional infrastructure determines pace. The cascade's speed and direction depend on patents, capital markets, education, regulation — not on the technology alone.
The transformative applications are not yet visible. The AI applications of 2026 are Newcomen's pump. The railroad equivalent — the combinatorial explosion that will define the era — lies decades in the future.