Pointing out that exponentials become sigmoids is not a forecast; Lindy’s Law sets the default expectation that AI scaling continues for roughly another seven years.
Key Takeaways
The sigmoid objection is technically true but proves nothing about when flattening occurs; UN birthrate and solar deployment forecasts both failed this same way repeatedly.
A Wharton team modeled the METR AI capabilities curve in early 2026 and predicted imminent flattening; the next model released immediately broke their projection.
If you model AI dynamics explicitly, you need projected data center growth, algorithmic progress rates, and engagement with existing work like the AI Futures Timeline Model.
If you treat AI as a black box, Lindy’s Law applies: scaling has run since roughly 2019, so median expectation is another ~seven years before a regime change.
Under a Pareto distribution assumption, the probability scaling ends within two more years is only about 22%.
Hacker News Comment Review
Thin discussion so far; the one comment gestures at a distinction between the 2017-2021 and 2022-2026 improvement drivers without elaborating, leaving the sigmoid-vs-exponential debate unresolved in the thread.
