Great read! Just a minor nitpick: did you mean _recurrent_ nets for RNNs rather than residual?