Bret Devereaux

Social organization does not scale up infinitely, especially before mass media (e.g. printing press, check out Imagined Communities). Consequently, manpower lies in effective heterogeneity in the ancient world.

This ends up being recursive. There is no point at which either the Mongols or the Romans are homogeneous. Rome's self-conception as a fusion-society goes back to its founding myths. Before the auxiliaries were diversity, the socii were, before that the Latins.

Chinggis Khan's tribal unit breaks with Jamukha and becomes independent for the first time precisely over Chinggis' willingness to recruit a wider and more heterogeneous mixture of mongols.

We could march this back into pre-history. Hat tip towards @MilHist_Lee , the 'killer app' of early humans was the ability to manage larger social groups and greater social complexity. Being able to have a tribal group that included members not blood-related.

Same goes for the development of the state - the ability to tolerate specialization and increasing heterogeneity gives the state the size-and-force advantage to expand. We don't see it because it is easy to miss because we're used to modern systems of social org...

Which are themselves the products of a multi-millennium arms race in social complexity, because the biggest-social-group typically wins. So we miss subtle group distinctions - Roman vs. Latin or Roman vs. Sabine - that actually mattered a lot to the people at the time.