The Exponential Decade: Navigating the 2020s When Technology Stopped Waiting

In September 2024, the measured rate of AI progress quietly changed - and the 2020s
stopped behaving like a normal decade.

The Exponential Decade is a field guide to the years we are actually living through,
built entirely from the documented record: the Epoch Capabilities Index, METR's task-

horizon benchmarks, lab-published safety frameworks, compute and energy projections,
and the dated forecasts of the people building the systems. It is a book about measured
curves, not marketing curves - and about what the slope change means for work, science,
and the next thirty-six months.

Inside, sourced and checkable:

- The two index projects that actually measure AI progress - and the September 2024
inflection both of them recorded
- Why AI task horizons are doubling every four months, measured in honest human-hours
- The episodic jumps: when frontier gains lurch forward, and when they regress to trend
- Compute, chips, and power: the real constraints behind the capability curve
- Timeline forecasts from Kurzweil's 2029 to lab-internal estimates - who published what,
and how the forecasts have aged
- The strong counterargument: the skeptics, the plateau case, and what would falsify each
- A closing chapter on what individuals, builders, and institutions can actually do with
the decade

Every chapter ends with key numbers, the skeptic's corner in full voice, and a dated
watchlist. For readers of Ray Kurzweil, Mustafa Suleyman, and anyone who wants the
numbers behind the headlines of the 2020s.

Chapter at a glance: The Shape of Exponential Change - The Frontier in 2026 -
Evidence of Recursive Improvement - Timeline Forecasts - Compute, Energy, and Constraints -
Beyond Benchmarks - The Counterargument - What To Do With the Decade