The empirical signature in the AI transition is visible in the capability curve. GPT-2 (February 2019) produced passable paragraphs that deteriorated into incoherence. GPT-3, sixteen months later, wrote essays and translated languages with startling fluency. GPT-4, thirty-three months after that, passed professional examinations. Claude 3.5 Sonnet, summer 2025, engaged in sustained context-sensitive collaboration on complex software. The interval between thresholds shrinks; the magnitude of each leap grows; the product is an acceleration curve that no linear projection can capture and no institutional planning cycle can track.
Toffler communicated the acceleration through his 800th lifetime argument: divide the last fifty thousand years of human existence into lifetimes of approximately sixty-two years each, and there are roughly eight hundred such lifetimes. Of those, six hundred and fifty were spent in caves. Writing has existed for only the last seventy. The printed word has reached masses for only the last six. Electric motors for only the last two. Most material goods in daily use have been developed within the present lifetime. The argument was designed to produce visceral recognition of the compression because the mind cannot process exponential curves intuitively.
The mind linearizes. It projects recent past into near future and assumes tomorrow resembles yesterday at roughly the same pace. This is not cognitive defect but evolutionary adaptation — a heuristic that served well for the first seven hundred and ninety-nine lifetimes, during which change was approximately linear. The heuristic fails catastrophically in the eight-hundredth lifetime, producing the specific phenomenon of temporal vertigo: the disorientation that occurs when the organism can no longer maintain a coherent sense of past-present-future relationships because the patterns that held last year no longer hold.
Segal captures temporal vertigo when he describes telling companies their 2026 planning, based on pre-December 2025 assumptions, was already obsolete. The plans were rational when made. The world they were made for had ceased to exist. The gap between plan and world, previously measured in years, opened in weeks. The executives were not slow. They were victims of a projection apparatus calibrated to a superseded rate of change.
Toffler's framework drew on mid-century systems theory, cybernetics (Wiener), and the economic history of innovation. He synthesized these into the claim that rate-of-change itself is the primary variable producing civilizational stress, independent of the content of any particular change.
The framework's relevance was disputed for decades on the grounds that Toffler's specific predictions were sometimes wrong. The AI transition has redirected the argument: whatever the status of specific predictions, the meta-prediction — that acceleration would eventually outrun adaptive capacity — has proven correct with a precision that even Toffler's defenders find alarming.
Second-order phenomenon. The acceleration is not change itself but the increase in the rate of change — a quantity invisible to organisms whose heuristics evolved for approximately linear environments.
Self-reinforcing loop. Products of innovation become tools of further innovation; the loop cannot be interrupted without dismantling the civilization producing it.
Ungovernable but navigable. The acceleration cannot be stopped, reversed, or wished away; what can be governed is the institutional environment through which it propagates to populations.
Temporal vertigo. When the past ceases to guide the present, projection apparatuses fail systematically, producing plans that are rational on their own terms and wrong about the world.
Pacing as policy. The pace at which AI disrupts existing structures can be influenced — not by restricting the technology but by governing the institutional environment within which it is deployed.