The cycle that began with [YOU] on AI is centered on the personal dimension of the AI transition: the individual who must decide whether and how to engage with tools that are rewriting what it means to work, create, and think. The exponential gap supplies the structural context for that personal decision. The individual who takes the orange pill and builds her capabilities with AI tools inhabits an institutional landscape that is almost certainly lagging the technology she is using. The legal frameworks governing her work, the labor protections available to her, the educational credentials that organized her career, and the regulatory environments constraining the companies whose tools she relies upon are all adapting to a world that the technology is actively transforming. Her personal adaptation is necessary but insufficient; the gap between what she can do and what her institutions have arranged for is not something she can close alone.
Azhar’s concept also explains a feature of the AI transition that observers repeatedly note without being able to account for: the coexistence of extraordinary technical capability with evident institutional dysfunction. The tools can do things that seem impossible. The organizations deploying them are struggling. The policy frameworks governing them are visibly inadequate. The exponential gap is not a description of a problem that might arise. It is a description of the condition in which the AI transition is occurring right now, in real time, as capability compounds and institutions walk.
Azhar derived the concept from the economic history of general-purpose technologies: the observation that electricity, the printing press, the steam engine, and the internal combustion engine each produced a period in which capability outran the institutional arrangements designed to govern it. In each case, the most consequential consequences arrived not with the headline invention but with the complementary inventions and organizational innovations that grew up around it — often decades later. The factory was reorganized not by the dynamo but by the electric motor, years after electrification began. Azhar’s insight was that this latency is not a historical accident but a structural feature: the institutional response to a general-purpose technology is inherently slower than the technology’s diffusion because the institutions must first understand what they are governing before they can govern it, and understanding arrives after the fact.
The concept gained its canonical formulation in his 2021 book and was quickly adopted by the policy community as the clearest available frame for understanding why AI governance felt perpetually behind. It also gained urgency from Azhar’s observation that AI is not merely another general-purpose technology but the convergence of several at once — computing, biology, energy, advanced manufacturing — each riding its own exponential curve, each compounding the others. The gap produced by one exponential technology is manageable. The gap produced by several simultaneous ones may not be.
The invisibility of the increment. The exponential gap is most dangerous because it is invisible in real time. Each year’s increment of technological change looks small and manageable. Observers keep judging the technology by the increment in front of them and missing the curve beneath their feet. The gap does not announce itself. It reveals itself all at once, when the accumulated increments have moved the world to a place the institutions were not designed to govern. By the time the gap is visible, the adaptive work is already years behind.
Structural mismatch, not personal failure. Azhar is careful to insist that the gap does not indict the people running the institutions. Regulators, courts, and corporations are doing exactly what they were built to do: move deliberately, demand evidence, balance interests, resist capture by any single faction. These are virtues in a linear world. Asking such institutions to keep pace with an exponential one is, as he puts it, like asking them to operate in a space they were never designed to enter. The failure is architectural, not personal.
Governance that can learn. The prescription Azhar draws from the diagnosis is governance that is itself adaptive — that updates as it goes rather than setting rules once and leaving them to ossify. The flat curve of regulation need not stay flat; it can be bent upward through institutional innovation that makes governance a continuous process of learning rather than a single act of rule-setting. The point is not to make institutions reckless but to make them faster learners, capable of revising judgments as evidence accumulates at something closer to the speed of the technology they govern.
The gap as the through-line. Every major theme in Azhar’s work — the economics of abundance, the concentration dynamics of networked markets, the dissolution of the social fictions that held industrial societies together, the geopolitical redistribution of power — is a manifestation of the same underlying structure: a technology that compresses distance faster than the institutions connecting people across it can adapt. The exponential gap is not a description of one problem. It is a description of the condition from which every other problem in the AI transition emerges.