Exponentials

Exponential growth occurs when a quantity increases by a consistent percentage over each time period rather than a consistent amount. The result is a curve that starts deceptively slow, then explodes. Human intuition is wired for linear extrapolation — we instinctively assume the next step will be about the same size as the last one. Exponential processes violate that assumption, which is why transformative technologies consistently surprise us: they seem like toys right up until the moment they reshape industries.

The Six Ds of Exponentials

Peter Diamandis has described the life cycle of exponential technologies through six stages — the Six Ds:

Digitization is the trigger. Once a technology or process becomes information-based — encoded as bits rather than atoms — it gains access to exponential improvement curves. Photography went from chemical film to digital sensors; money went from paper to electronic ledgers; intelligence is going from human brains to neural networks. Digitization is the on-ramp to everything that follows.

Deception is the long flat stretch at the start of the exponential curve. Doubling from 0.01 to 0.02 to 0.04 looks like nothing is happening. AI spent decades in this phase — researchers made incremental progress that seemed academic and impractical. 3D rendering was a curiosity in the 1980s. VR was a punchline through the 2010s. The deceptive phase is where most observers conclude the technology "doesn't work" and move on, missing that doublings are about to enter visible territory.

Disruption is the moment the exponential curve hits a threshold where the technology becomes competitive with — then superior to — incumbent solutions. When AI models crossed from "interesting research" to "better than most humans at specific tasks" between 2022 and 2024, the disruption phase began. When real-time ray tracing went from impossible to standard on consumer GPUs (Huang's Law), 3D rendering was disrupted. Disruption isn't gradual — it feels sudden because the exponential curve went from imperceptible to overwhelming in a few doublings.

Demonetization: as exponential technologies mature, they strip the cost out of existing markets. The cost of AI inference has fallen roughly 92% since early 2023. The cost of genome sequencing fell from $3 billion (Human Genome Project) to under $200. The cost of solar energy per watt has followed Wright's Law curves for decades. Demonetization is Deflationary Technology in action — the technology becomes so cheap that the old business model built on scarcity collapses.

Dematerialization: products disappear into software. The smartphone dematerialized the camera, GPS unit, music player, flashlight, compass, calculator, voice recorder, alarm clock, and map. AI is dematerializing the junior analyst, the customer service agent, and the first-draft writer. In the metaverse, digital twins dematerialize physical prototyping — you simulate the factory before building it.

Democratization: once demonetized and dematerialized, the technology becomes accessible to everyone. Tools that were once the exclusive domain of well-funded organizations reach individuals. AI coding tools democratize software development — the transition from the Engineering Era to the Creator Era that defines the creator economy. Game engines democratize 3D world-building. Open-weight foundation models democratize AI capability itself. Democratization is where Jevons' Paradox fully ignites: when a billion people can do what previously required a specialized team, total usage explodes.

Exponential Curves That Define the Agentic Economy

Several exponential curves are converging simultaneously in the agentic economy. Huang's Law drives GPU performance for AI workloads faster than Moore's Law ever drove CPUs. Wright's Law drives the cost of AI inference down a predictable curve with each doubling of cumulative compute deployed. Reed's Law drives the value of interconnected agent networks exponentially as more agents join the ecosystem. And the data flywheel — more users generate more data, which improves models, which attract more users — creates its own exponential loop.

The reason exponential processes are so hard to plan for is that human institutions — regulation, education, corporate strategy, infrastructure planning — operate on linear timescales. A regulatory framework designed for AI capabilities as they existed in 2024 is already inadequate for 2026, and will be comically insufficient by 2028. Companies that plan based on the current state of AI rather than its exponential trajectory are making the same mistake as Kodak planning around film in 2005. The organizations that thrive in exponential environments are those that plan for the curve, not the current position on it.