We describe a class of asymptotically good codes built from the intersection of randomly permuted binary BCH codes. This family of pseudo-random error correcting codes, called Generalized Low Density (GLD) codes, is a direct generalization of Gallager's Low Density Parity Check (LDPC) codes. GLD codes belong to the larger family of Tanner codes based on a random bipartite graph. We study the GLD ensemble performance and prove the asymptotically good property. We also compare GLD codes minimum distance and performance to the Varshamov-Gilbert bound and BSC capacity respectively. The results show that Maximum-Likelihood decoding of GLD codes achieves near capacity efficiency. The suboptimal iterative decoding of GLD codes is briefly presented. Experimental results of small and large blocklength codes are finally illustrated on both AWGN and Rayleigh fading channels.