The sigmoids won't save you

# Summary The article argues that while the common AI talking point "all exponentials eventually become sigmoids" (meaning growth trends flatten out) is technically true, it's often misapplied to prematurely predict when AI capabilities will plateau. Scott Alexander demonstrates through examples like UN birthrate projections, solar power deployment, and AI capability models that forecasters frequently predict sigmoid flattening too early, only to see exponential growth continue longer than expected. The key insight is that understanding the underlying mechanisms driving growth—rather than simply assuming a curve must flatten—is necessary to accurately predict when exponential trends actually end.