This paper introduces a self-supervised approach for learning node and graph-level representations by contrasting structural views of graphs. The method contrasts encodings from first-order neighbors and a graph diffusion, achieving state-of-the-art results on 8 out of 8 node and graph classification benchmarks under the linear evaluation protocol. On the Cora node classification benchmark, the approach achieves 86.8% accuracy, a 5.5% relative improvement over previous state-of-the-art. On the Reddit-Binary graph classification benchmark, it achieves 84.5% accuracy, a 2.4% relative improvement over previous state-of-the-art. The method outperforms supervised baselines on 4 out of 8 benchmarks. The approach uses two structural views of graphs: first-order neighbors and a graph diffusion. It employs a graph diffusion to generate an additional structural view of a sample graph, which is then sub-sampled and fed to two dedicated GNNs followed by a shared MLP to learn node representations. The learned features are then fed to a graph pooling layer followed by a shared MLP to learn graph representations. A discriminator contrasts node representations from one view with graph representation from another view and vice versa, and scores the agreement between representations which is used as the training signal. The method is evaluated on three node classification and five graph classification benchmarks. It achieves state-of-the-art results on both node and graph classification tasks. The results show that contrasting node and graph encodings across views achieves better results on both tasks compared to contrasting graph-graph or multi-scale encodings. A simple graph readout layer achieves better performance on both tasks compared to hierarchical graph pooling methods such as differentiable pooling. Applying regularization or normalization layers has a negative effect on the performance. The method is compared with state-of-the-art models, including unsupervised models such as DeepWalk and DGI, and supervised models such as MLP, ICA, LP, ManiReg, SemiEmb, Planetoid, Chebyshev, MoNet, JKNet, GCN, and GAT. The results show that the method achieves state-of-the-art results on both node and graph classification tasks. It outperforms supervised baselines on 4 out of 8 benchmarks. The method is also compared with graph kernel methods and unsupervised methods, achieving state-of-the-art results on both. The method is shown to be effective in both node and graph classification tasks without requiring specialized architectures.This paper introduces a self-supervised approach for learning node and graph-level representations by contrasting structural views of graphs. The method contrasts encodings from first-order neighbors and a graph diffusion, achieving state-of-the-art results on 8 out of 8 node and graph classification benchmarks under the linear evaluation protocol. On the Cora node classification benchmark, the approach achieves 86.8% accuracy, a 5.5% relative improvement over previous state-of-the-art. On the Reddit-Binary graph classification benchmark, it achieves 84.5% accuracy, a 2.4% relative improvement over previous state-of-the-art. The method outperforms supervised baselines on 4 out of 8 benchmarks. The approach uses two structural views of graphs: first-order neighbors and a graph diffusion. It employs a graph diffusion to generate an additional structural view of a sample graph, which is then sub-sampled and fed to two dedicated GNNs followed by a shared MLP to learn node representations. The learned features are then fed to a graph pooling layer followed by a shared MLP to learn graph representations. A discriminator contrasts node representations from one view with graph representation from another view and vice versa, and scores the agreement between representations which is used as the training signal. The method is evaluated on three node classification and five graph classification benchmarks. It achieves state-of-the-art results on both node and graph classification tasks. The results show that contrasting node and graph encodings across views achieves better results on both tasks compared to contrasting graph-graph or multi-scale encodings. A simple graph readout layer achieves better performance on both tasks compared to hierarchical graph pooling methods such as differentiable pooling. Applying regularization or normalization layers has a negative effect on the performance. The method is compared with state-of-the-art models, including unsupervised models such as DeepWalk and DGI, and supervised models such as MLP, ICA, LP, ManiReg, SemiEmb, Planetoid, Chebyshev, MoNet, JKNet, GCN, and GAT. The results show that the method achieves state-of-the-art results on both node and graph classification tasks. It outperforms supervised baselines on 4 out of 8 benchmarks. The method is also compared with graph kernel methods and unsupervised methods, achieving state-of-the-art results on both. The method is shown to be effective in both node and graph classification tasks without requiring specialized architectures.