Thousands of cell-type-specific split-GAL4 driver lines and a massive, searchable image collection enable Drosophila neuroscientists to target neurons of interest with increased precision.

In feedforward networks with random connectivity, connection motifs underlying the detection of short clusters and activity sequences are likely to converge on a subset of postsynaptic neurons purely by chance.

The dynamical solutions exhibited by task-optimized recurrent neural networks, and their similarity to prefrontal cortex dynamics, depends strongly on the strength of neural noise and metabolic cost imposed during training.