Phenomenology of the Exponential
In The You On AI Field Guide
Intel's 4004 microprocessor (1971) contained 2,300 transistors. The 8080 (1974) contained 4,500 — a doubling. A meaningful engineering achievement, nothing that felt like revolution to the engineers who accomplished it. The 8086 (1978) contained 29,000. The 80286 (1982) contained 134,000. Each generation represented roughly a doubling; each felt, from inside, like the natural next step. The engineers were methodical, not astonished. They had a roadmap; they followed it. The cumulative effect of following that roadmap for five decades is a modern processor containing tens of billions of transistors — more computational power than all computers existing on the planet when Moore wrote his 1965 paper.
The people most surprised by the AI transition of 2025 were, paradoxically, the people who had lived through every previous increment. They were the experienced technologists who had watched each interface transition arrive and concluded, reasonably, that they understood the trajectory. They had been on the chessboard for decades. What they did not grasp is where on the chessboard they were sitting. The first half is manageable: each doubling adds quantity within human intuition. The second half exceeds any reference frame the mind can supply. The AI transition sits squarely on the second half of the computational chessboard.
Consider the numbers: total compute used to train AI models has doubled approximately every six months since 2012, far faster than Moore's Law's two-year cadence. Stanford researchers measured the post-2012 doubling time at 3.4 months for the most compute-intensive runs. Jensen Huang described the progression as 'Moore's Law squared.' By 2025, computational power applied to training a single frontier model exceeded what the entire semiconductor industry produced annually in the decade when Moore first made his observation. These numbers are not secrets — but they are second-half-of-the-chessboard numbers, and stating them does not convey their implications.
Moore's career provides a corrective to this cognitive failure: respect the trend, even when it exceeds intuition. In engineering, the discipline takes the form of the roadmap. The International Technology Roadmap for Semiconductors existed precisely because engineers could not intuit exponential trajectories — they needed the roadmap to organize the present around a trajectory too steep for individual intuition. AI has no equivalent coordinated roadmap, which is closer to the early semiconductor industry before coordination emerged. Moore's experience suggests the transition from competitive chaos to coordinated roadmap is inevitable for any technology on an exponential curve, because the economics of each doubling eventually exceed what single companies can bear.
Origin
The phenomenological observation is as old as exponential mathematics — the chessboard parable dates to medieval Persia — but its application to technology management emerged through the semiconductor industry's hard-won discipline of roadmap-based planning. Moore discussed the pattern implicitly across multiple interviews, noting that intuitions calibrated to linear progress systematically mislead on exponential curves.
Key Ideas
Each doubling feels incremental. The subjective experience of exponential growth is linear; only the cumulative effect, viewed in retrospect, reveals the exponential character.
The chessboard has two halves. The first half is within human intuition; the second half exceeds it. The AI transition sits on the second half.
Roadmaps are cognitive prosthetics. The semiconductor industry's coordinated planning existed precisely because engineers could not intuit the trajectory.
Surprise is a failure of intuition. The people closest to the curve are often the most surprised by its consequences, because incremental familiarity obscures cumulative magnitude.
Preparation, not astonishment. The appropriate engineering response is to plan for the next doubling, not to marvel at the last one.