Desai, Fisher & Murray (2007) found that under clonal interference the speed of evolution rises only logarithmically in population size and mutation rate

population-geneticsclonal-interferenceevolutionary-ratesdiminishing-returnsasexual-populationsscaling-laws

"The speed of evolution and maintenance of variation in asexual populations" (Michael M. Desai, Daniel S. Fisher & Andrew W. Murray, Current Biology 17:5, 2007) analyzes what happens when a large asexual population generates beneficial mutations faster than any one of them can fix. Multiple favorable lineages then compete simultaneously — clonal interference — and most are lost even though they are advantageous, because a fitter competitor sweeps first. The consequence is a hard limit on how fast adaptation can proceed: "the speed of evolution increases only as the logarithm of the population size and the logarithm of the mutation rate."

Logarithmic dependence is a severe diminishing return. Multiplying the population by a factor of ten does not multiply the adaptation rate by ten; it adds a roughly constant increment. The extra individuals overwhelmingly supply mutations that interfere with, rather than accelerate, the fixation of the best variant. The result extends the Gerrish & Lenski (1998) clonal-interference lineage into an analytic scaling law.

This is the biology instance of the same brake economics found in claim-bloom-2020-ideas-are-getting-harder-to-find. It closes the loop on the population-as-engine story: the acceleration mechanism of claim-kremer-1993-technology-growth-proportional-to-population and claim-hawks-2007-human-adaptive-evolution-accelerated-recently — more agents, more variation, faster improvement — is not free in either domain. Beyond a point, scaling the generating population yields sub-linear returns, and the friction is not an economics artifact but a general property of populations generating variation. The cross-domain pattern is drawn together in observation-population-scaled-improvement-hits-a-sublinear-brake-across-domains.