Show HN: New compression algorithm beats ZSTD by 14%

Hi all, I made a data compression algorithm that works as follows: Replace common bigrams with a fixed value: "of the" => "of e" "and she" => "and e" "when there" => "when e"

To decompress, look up the bigram dictionary for the right value of e.

I've had some weird bugs in the decompression process, so I'm looking forward to others looking at this.