Extrapolating data from 90 years of an isolated island nation to evolutionary timescales is the major assumption. The other weak foundation is the assumption that this gene should correspond to educational level. It was weakly correlated in the early 20th century, less correlated now, and we can't put any weight on to what it actually controls (maybe it's related to IQ, maybe it's delaying gratification, who knows, the writers don't claim to).

I think it's a bold conclusion to leap from one genetic marker doesn't seem as strongly correlated to education levels among Icelanders to "cognitive genomics is on the decline".