Cortex codes are short length block codes having a large Hamming distance. Their modular construction, based on simple and very short block codes, yield to difficulties in efficiently decoding them with digital decoders implementing the Sum-Product algorithm. However, this construction lends itself to analog decoding with performance close to ML decoding as was recently demonstrated. A digital decoding method close to analog decoding is stochastic decoding. This paper brings the two together to design a Cortex stochastic architecture with good decoding performance. Moreover, the proposed stochastic decoder architecture is simplified when compared to the customary one. Instead of edge or tracking forecast memories the proposed architecture uses multiple streams to represent the same probability and deterministic shufflers. This results in a more efficient architecture in terms of ratio between data throughput and hardware complexity. Finally, the proposed method offers decoding performance similar to a Min-Sum decoder with 50 iterations.