Excellent speech.
I think we should view market capitalism as like an AI algorithm in a very convoluted space.
Start everyone off with the same amount of resources, and then let market capitalism go. For the first few iterations, inequality is forming but everyone is better off; even the poor are making absolute gains. You'll conclude that the algorithm (market capitalism) works. Untended, though, it starts to diverge after some time. The parameters go to infinity, the model gets ridiculous and ceases to represent real information (which institutions are worthy and add value, which do not) and goes into the self-perpetuation of early, parochial advantage (like overtraining, in ML, where noise in the data is misinterpreted as genuine signal).
What we want is to have the growth of that initial training arc, not the degeneracy as the algorithm diverges, and you want to keep having that growth and increasing general prosperity indefinitely. So we need continuous regularization. The welfare state is supposed to do that, but the increasing ability for those with wealth to convert it into power has rendered our government severely dysfunctional, so the normalization has stopped.
The problem with poverty in capitalism is that there's no real solution. Divergence and perpetuation of poverty is exactly what you'd expect. Markets are all about trade, but a person with nothing-- no social capital, no cultural capital, no wealth-- has nothing to trade, except for when terms for labor are extremely favorable-- like, skilled labor is in such shortage that even the out-of-touch rich see value in training the poor, for free-- and that's something we haven't seen since the 1960s.