Large good codes were always known to exist, the crux from when Shannon published his work (I believe) was to actually get good codes that are computationally v

Large good codes were always known to exist, the crux from when Shannon published his work (I believe) was to actually get good codes that are computationally viable (else you could even use Shannon's random code). LDPC codes were rediscovered only because decoding became viable due increased computational power (w.r.t the 60's) and, crucially, discovery of iterative decoding for them (belief propagation etc).