Causal Relation Network (CausalRN) is an MLP sequence modeling architecture that can match Transformers on the copying task, employing key innovations such as exponentially-activated RNs, pre-activation normalization, and providing insights into strong in-context retrieval.