It's not always a joke. There's dozens if not hundreds of people who believe they're cracked the compression problem by "just finding the offset of your data into pi". It is a recurring problem online: https://duckduckgo.com/?q=compress+pi+offset (DDG actually did better than Google here which is why I link that), and sometimes explaining to the proposer why this doesn't work is quite difficult. Sometimes you end up reduced to "Well, just implement it then if it's so easy and amazing."

(Fortunately, pi compression is not particularly amenable to one of the other basic "tricks" of fake compression algorithms, which is failing to count filenames or other metadata as part of the compressed data set, if you use them in the decompression algorithm. Pi "compression" is, in practice, a much better data expansion algorithm than a decompression algorithm. So you don't generally have to deal with having to explain how they've still failed even though their file sizes seem smaller.)