The persistent claim that AI growth curves must flatten like sigmoids ignores a crucial reality: exponential trends often persist far longer than forecasters expect. While technically true that all growth eventually hits limits, the timing matters enormously for practical planning.

Historical examples demonstrate this misprediction pattern repeatedly. UN birthrate projections consistently assumed flattening that never arrived. Solar power deployment forecasts similarly underestimated sustained exponential growth. Most notably, Wharton researchers predicted AI capability curves would plateau, only to watch them explode beyond their projections.

The solution lies in Lindy's Law: without deep process understanding, assume trends continue as long as they've already persisted. Since dramatic AI improvement began around 2017-2019, this suggests several more years of rapid scaling. Critics claiming AI won't reach concerning capability levels must justify their skepticism with specific models, not hand-waving about natural limits.

Burden of proof matters here. Anyone arguing AI capabilities won't escalate must explain their modeling assumptions and address existing research like the AI Futures Timeline Model. Default assumptions should favor continued exponential growth rather than premature flattening.